ON THE CONTINGENT VALUE OF DYNAMIC CAPABILITIES FOR … · 2014. 1. 14. · on the relationship between dynamic capabilities, environmental dynamism, and competitive advan-tage. Integrating

Strategic Management JournalStrat. Mgmt. J., 35: 179–203 (2014)

Published online EarlyView 6 May 2013 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smj.2099

Received 22 January 2012 ; Final revision received 18 January 2013

ON THE CONTINGENT VALUE OF DYNAMICCAPABILITIES FOR COMPETITIVE ADVANTAGE:THE NONLINEAR MODERATING EFFECT OFENVIRONMENTAL DYNAMISMOLIVER SCHILKE*Department of Sociology, University of California, Los Angeles, California, U.S.A.

This article suggests that dynamic capabilities can give the firm competitive advantage, but thiseffect is contingent on the level of dynamism of the firm’s external environment. A nonlinear,inverse U-shaped moderation is proposed, implying that the relationship between dynamiccapabilities and competitive advantage is strongest under intermediate levels of dynamism butcomparatively weaker when dynamism is low or high. This proposition is tested using data onalliance management capability and new product development capability, two specific dynamiccapabilities widely recognized in prior research. Results based on longitudinal key informantdata from 279 firms support the account that these dynamic capabilities are more stronglyassociated with competitive advantage in moderately dynamic than in stable or highly dynamicenvironments. Copyright 2013 John Wiley & Sons, Ltd.

INTRODUCTION

The dynamic capabilities perspective has emergedas one of the most influential theoretical lensesin the study of strategic management over thepast decade. Despite its popularity in the litera-ture, the dynamic capabilities perspective has beencriticized for its ill-defined boundary conditionsand its confounding discussion of the effect ofdynamic capabilities (e.g., Arend and Bromiley,2009). One important source of concern is thatthe presence of dynamic capabilities has frequentlybeen equated with environmental conditions char-acterized by high dynamism (Zahra, Sapienza,and Davidsson, 2006). A turbulent environment,

Keywords: dynamic capabilities; contingency perspec-tive; competitive advantage; environmental dynamism;survey research*Correspondence to: Oliver Schilke, Department of Sociology,University of California, Los Angeles, 264 Haines Hall, 375Portola Plaza, Los Angeles, CA 90095–1551, U.S.A. E-mail:[email protected]

Copyright 2013 John Wiley & Sons, Ltd.

however, is not necessarily a component or pre-condition of dynamic capabilities, which can existeven in stable environments (Helfat and Winter,2011). Further, researchers have tended to iden-tify dynamic capabilities post hoc, often equatingtheir existence with successful organizational out-comes. This practice makes it difficult to separatethe existence of dynamic capabilities from theireffects (Helfat and Peteraf, 2009). Given the abovelimitations, it has remained difficult to ascertain thevalue of dynamic capabilities for a firm’s compet-itive advantage, especially under different degreesof dynamism.

This article empirically investigates the linkbetween dynamic capabilities and competitiveadvantage and examines the efficacy of dynamiccapabilities under conditions of varying environ-mental dynamism. To accomplish this goal, I con-ceptualize dynamic capabilities in terms of orga-nizational routines, thus making them measurableand distinct from a firm’s competitive advantage(Eisenhardt and Martin, 2000). I also separate

180 O. Schilke

dynamic capabilities from the firm’s external envi-ronment, which I identify and measure as a con-tingency factor (Helfat et al., 2007). Making thisdistinction allows for considering and ultimatelyreconciling the competing claims regarding theeffect of environmental dynamism on the relation-ship between dynamic capabilities and competi-tive advantage. Some propose that the dynamismof a firm’s environment may enhance the effi-cacy of dynamic capabilities and their potentialfor competitive advantage (Drnevich and Kriauci-unas, 2011; Winter, 2003; Zollo and Winter, 2002).Other scholars, however, suggest that, on the con-trary, dynamic capabilities may prove less effec-tive in highly dynamic environments (Eisenhardtand Martin, 2000; Schreyögg and Kliesch-Eberl,2007).

The key contributions of this research aretwofold. First, this article makes a theoretical con-tribution by offering a new, integrative positionon the relationship between dynamic capabilities,environmental dynamism, and competitive advan-tage. Integrating existing views, I propose a novelinverse U-shaped moderating effect, implying thatthe association between dynamic capabilities andcompetitive advantage is strongest under inter-mediate levels of dynamism but comparativelyweaker when dynamism is low or high. Second,this study makes an empirical contribution by test-ing this nonlinear interaction effect. In doing so,this article contributes to reducing the scarcityof empirical research on the consequences ofdynamic capabilities for organizational outcomes(e.g., Helfat and Peteraf, 2009).

THEORETICAL BACKGROUND

Dynamic capabilities and competitiveadvantage

The dynamic capabilities view may be regarded asan extension of the resource-based view (RBV);while the RBV primarily addresses a firm’sexisting resources, the dynamic capabilities viewemphasizes the reconfiguration of these resources(Helfat and Peteraf, 2003). Prior research sug-gests that dynamic capabilities are organiza-tional routines that affect change in the firm’sexisting resource base (Eisenhardt and Martin,2000; Helfat, 1997; Teece, Pisano, and Shuen,1997). This definition emphasizes that dynamic

capabilities are based on organizational rou-tines, commonly understood as learned, highlypatterned, repetitious behavioral patterns forinterdependent corporate actions (Pierce, Boerner,and Teece, 2002; Winter, 2003; Zollo and Winter,2002). Although the routines underlying dynamiccapabilities are not entirely fixed as people per-form them across time and space, interpret themsubjectively, and ultimately introduce variations(Feldman and Pentland, 2003), Winter (2003)emphasizes that there are clear limits to the degreeto which they reflect flexible action with modestcontinuity across occasions (also see Weick andSutcliffe, 2006). Therefore, it is important tonote that not all organizational change needs tooriginate from dynamic capabilities; in particular,contingent, creative improvisation is typically notassociated with dynamic capabilities as definedhere (Winter, 2003).

Interest in dynamic capabilities stems from theirpotential influence on competitive advantage, thekey outcome variable in dynamic capabilities the-ory (Teece et al., 1997). A firm is said to have acompetitive advantage when it enjoys greater suc-cess than current or potential competitors in itsindustry (Peteraf and Barney, 2003). Consistentwith this conceptualization, superior firm perfor-mance relative to rivals commonly serves as anempirical indicator of competitive advantage.

Traditionally, the literature has assumed auniversally positive effect of dynamic capabilitieson competitive advantage. By replacing exist-ing resources, dynamic capabilities have beensuggested to create better matches between theconfiguration of a firm’s resources and externalenvironmental conditions (e.g., Teece and Pisano,1994).

The contingent role of environmentaldynamism

However, researchers have started to disagreein their assessments of the value of dynamiccapabilities. Advocates of a more contingent viewposit that the benefits of dynamic capabilitiesdepend not only on the existence of the underlyingorganizational routines, but also on the context inwhich these capabilities are deployed (Levinthal,2000; Sirmon and Hitt, 2009). Recognizing thateffective modes of organizational adaptation areat least partly determined by environmental forces(Hrebiniak and Joyce, 1985), recent theoretical

Copyright 2013 John Wiley & Sons, Ltd. Strat. Mgmt. J., 35: 179–203 (2014)DOI: 10.1002/smj

On the Contingent Value of Dynamic Capabilities 181

accounts on dynamic capabilities have emphasizedparticularly the role of environmental dynamism asa potentially important contextual variable (Helfatet al., 2007; Helfat and Winter, 2011; Zahra et al.,2006).

In defining environmental dynamism, thisarticle builds on Miller and Friesen’s (1983)influential conception that views both volatility(rate and amount of change) and unpredictability(uncertainty) as fundamental characteristics ofenvironmental dynamism. For example, changesin industry structure, the instability of marketdemand, and the probability of environmentalshocks are important elements of environmentaldynamism (e.g., Jansen, Van Den Bosch, and Vol-berda, 2006; Levinthal and Myatt, 1994; Sirmon,Hitt, and Ireland, 2007). Consequently, envi-ronments with little dynamism are characterizedby infrequent changes, and market participantsusually anticipate those changes that do occur.In contrast, highly dynamic environments arethose where rapid and discontinuous changes arecommon. In the middle lie moderately dynamicenvironments with regular changes that occuralong roughly predictable and linear paths.

Currently, there are two competing views onthe effect of environmental dynamism on thelink between dynamic capabilities and competitiveadvantage, with little integration of both perspec-tives. The first view posits that there has to bea critical need to change in order to gain sig-nificant value from these capabilities (Drnevichand Kriauciunas, 2011; Helfat et al., 2007; Win-ter, 2003; Zahra et al., 2006; Zollo and Winter,2002). This is because building and using dynamiccapabilities are costly. These costs typically arisefrom various activities involved in devising newresources, reconfiguring existing ones, and com-binations thereof. Additional costs might accrueif continual reconfigurations of resources unneces-sarily disrupt ongoing learning activities by pre-venting the firm from recognizing potential differ-ences in the outcome of its resources under differ-ent conditions. Other significant costs may resultfrom wrongly estimating the need for resourcealterations, which happens when firms use theirdynamic capabilities when there is no compellingneed for change (Winter, 2003). This can cre-ate significant costs because the frequent disrup-tion of the underlying resource base may degradestructural reproducibility and hence decrease an

organization’s ability to act as a reliable andaccountable collective entity.

Clearly, acknowledging that developing dy-namic capabilities involves serious costs has impli-cations for their potential value. If a firm rarelyhas a need to change, its performance relative tocompetitors may suffer when it devotes significantresources to developing these capabilities. Thisobservation emphasizes the importance of balanc-ing the costs of a given dynamic capability and itsactual use. As such, dynamic capabilities can beviewed as ‘strategic options’ (Kogut and Zander,1996) that allow firms to (re)shape their existingresource base when the opportunity or need arises.The lower the need for change, the less likelythe opportunity to ‘strike’ the option, makingdynamic capabilities comparatively less valuable.This implies that a firm needs to use its dynamiccapabilities repeatedly in order for them to producesignificant value (Helfat and Winter, 2011).

Following this logic, in environments character-ized by low dynamism, dynamic capabilities canbe expected to be of relatively less importancefor a firm’s competitive advantage. These environ-ments typically reward consistent exploitation ofexisting resources (Leonard-Barton, 1992; Teece,2007), whereas constantly reconfiguring resourcesmay disrupt the efficiency and value potential ofthe firm’s resources. Consequently, the positiveeffect of dynamic capabilities on a firm’s compet-itive advantage will be comparatively low whenenvironmental dynamism is low.

Another group of researchers has stressed thatroutine-based dynamic capabilities are not alwaysan adequate means of change, even if there is asignificant need for resource configurations (Eisen-hardt and Martin, 2000; Schreyögg and Kliesch-Eberl, 2007). An important feature of the routinesunderlying dynamic capabilities is that they arepath dependent and therefore based on interpre-tations and outcomes of past actions (Schreyöggand Kliesch-Eberl, 2007). Routine-based, history-dependent organizational change is typically veryeffective for adapting locally and incrementallybased on past experiences, but research on expe-riential learning argues that this type of orga-nizational change may prove problematic whenpreviously unknown forces continuously alter thebasis of competitive success (Levinthal and Rerup,2006; March and Levinthal, 1993), as is the casein highly dynamic environments. More specif-ically, contexts where change is frequent and


182 O. Schilke

unpredictable and the environment shifts uncer-tainly among states that place novel demands onthe organization produce two kinds of problemsfor dynamic capabilities, the first of which I calla ‘matching problem’ and the second an ‘inertiaproblem.’

The matching problem is intrinsic to the wayin which dynamic capabilities work. Following apatterned stimulus–response logic, they match par-ticular environmental states with certain avenuesfor organizational change (Levinthal, 2000; Pierceet al., 2002). For this purpose, the environment ismonitored, and appropriate organizational changesthat proved successful under similar conditions inthe past are invoked (Levinthal and Rerup, 2006;March and Levinthal, 1993). For this experience-based matching process to work, however, theorganization must have encountered the particu-lar (or at least a comparable) environmental statebefore. In Weick and Sutcliffe’s (2006) terminol-ogy, the environment needs to be in an ‘in-family’state—a situation that was previously experienced,analyzed, and understood. ‘Out-of-family’ states,which are common when environmental dynamismis high, pose problems to the effectiveness ofdynamic capabilities in that they do not triggera programmed reactivation of matching organiza-tional change. Given the absence of relevant stim-ulus knowledge, an out-of-family state may eitherbe ignored or become normalized—that is, treatedas if it were a familiar event already encounteredand understood in the past—, and potentially inap-propriate organizational responses may in turn bematched to these normalized situations (Levinthaland Rerup, 2006; Weick and Sutcliffe, 2006).

Second, even when environmental states appearfamiliar and previously successful organizationalresponses can be identified, this does not neces-sarily ensure that the same response will againand again be the most effective one (Jansenet al., 2006; Pierce et al., 2002). The automatedstimulus–response logic underlying dynamic capa-bilities, however, tends not to incentivize scrutiny.Given that a proven response to an identifiedproblem exists in organizational memory, exper-imentation with alternatives becomes less attrac-tive, crowding out explorative activities that wouldgo beyond the beaten track (Levinthal, 1991;Levinthal and Myatt, 1994; Sørensen and Stuart,2000). This issue is what I call an inertia problem.Importantly, it can pertain not only to an organi-zation’s zero-order capabilities but to its dynamic

capabilities as well, since routine-based organi-zational change tends to favor local adaptations(Collis, 1994; Levinthal, 1997; Schreyögg andKliesch-Eberl, 2007). Especially when environ-mental dynamism is high and contextual changeis fundamental and discontinuous, long-jump reori-entations that require entirely novel solutions oftenprove more beneficial for a firm’s competitiveadvantage than local adaptations from within thecurrent set of available actions (Levinthal, 1997,2000; Sørensen and Stuart, 2000).

In sum, I propose that highly dynamic environ-ments with their unfamiliar states and demandfor novel actions pose distinct challenges to theeffectiveness of dynamic capabilities. Matchingunfamiliar situations with organizational changesproves difficult and may lead either to unrespon-siveness or normalization and, in turn, implemen-tation of inappropriate responses. Additionally,experience-based adaptation is often associatedwith inertial forces that impede employing lesslocal, more path-breaking changes that are oftenrequired for organizations in highly dynamicenvironments to create a competitive advantage.

Overall, I recognize that environmental dy-namism affects both the extent of opportunities tochange and the organizational capacity to exploitthese opportunities via routine-based change, thusacknowledging the validity of the argumentsfrom both research camps. When environmentaldynamism is low , the potential of dynamic capabil-ities is limited because there are few occasions toexercise them effectively. In these situations, orga-nizational routines for adapting the resource basemay be of reduced value, in particular when con-sidering the costs associated with them. Therefore,when environmental dynamism is low, I suggestthat dynamic capabilities exert a relatively weakinfluence on the competitive advantage of firms.

I expect that when environmental dynamism ishigh , dynamic capabilities may also have a rela-tively weak impact on the competitive advantageof firms. Although highly dynamic environmentsprovide ample opportunities for resource reconfig-urations, the high frequency of novel situations andthe necessity to bring about discontinuous organi-zational change in these settings make the routine-based mechanisms dynamic capabilities rest oncomparatively less appropriate, given the matchingand inertia problems associated with them.

In contrast, I expect that dynamic capabilitieshave the relatively strongest positive effect on the



competitive advantage of firms when environmen-tal dynamism is intermediate. These environmentsare dynamic enough to create opportunities forchange but stable enough for organizations to rec-ognize reoccurring problem structures and success-fully leverage solutions existing in organizationalmemory. When environmental dynamism is at theintermediate level, there is both a potential fororganizational change and the capacity to makegood use of the routinized practices that under-lie dynamic capabilities. In summary, I expectthe positive effect of dynamic capabilities first toincrease but then to diminish as environmentaldynamism continues to rise, eventually decliningat high levels of dynamism. I test this positionempirically below.

HYPOTHESES

Dynamic capabilities manifest themselves in var-ious identifiable and specific business processes(Eisenhardt and Martin, 2000; Helfat et al., 2007;Helfat and Winter, 2011). Thus, rather than mea-suring a necessarily vague, generic dynamic capa-bility, empirical researchers have been advised tocarefully select a set of relevant business pro-cesses in which these capabilities exist to testtheir hypotheses (Gruber et al., 2010; Helfat andPeteraf, 2009; Helfat and Winter, 2011). Althoughselecting a limited number of specific processesas proxies for dynamic capabilities may affectthe universality of results, doing so is necessaryfor empirical research on dynamic capabilities tobe practicable. It is through theoretical inductionthat such empirical research on specific types ofdynamic capabilities ‘sheds light not only on thesespecific processes, but also on the generalizednature of dynamic capabilities’ (Eisenhardt andMartin, 2000: 1108).

In this study, I develop and test hypotheses onthe contingent dynamic capabilities–competitiveadvantage link using data on alliance manage-ment capability and new product developmentcapability . I selected these two dynamic capa-bilities for various related reasons. First, strate-gic alliances and new product development areessential means for reconfiguring the organiza-tional resource base. While strategic alliances givefirms access to resources that lie outside of theirboundaries (Das and Teng, 2000), new productdevelopment aims at updating the firm’s product

portfolio (Helfat and Raubitschek, 2000). Second,existing definitions of both alliance managementcapability and new product development capabil-ity are a good match with the conceptualizationof dynamic capabilities adopted here. Helfat et al.(2007: 66) define alliance management capabilityas a ‘type of dynamic capability with the capac-ity to purposefully create, extend, or modify thefirm’s resource base, augmented to include theresources of its alliance partners’ (see also Schilkeand Goerzen, 2010). New product developmentcapability is commonly defined as organizationalroutines that purposefully reconfigure the organiza-tional product portfolio (Danneels, 2008; Lawsonand Samson, 2001; Subramaniam and Venkatra-man, 2001). Third, alliance management capabil-ity and new product development capability areamong the most frequently mentioned types ofdynamic capabilities in the extant literature (e.g.,Eisenhardt and Martin, 2000; Helfat et al., 2007;Helfat and Winter, 2011; Teece and Pisano, 1994).Fourth, in the explorative fieldwork, alliance man-agement and new product development were themost frequently named types of routine activitiesfor adapting organizations to changes in the envi-ronment (see the Method section). Taken together,these two capabilities are particularly representa-tive for the dynamic capabilities concept, whichmakes them ideal candidates for this study.

In what follows, I develop two hypotheses forthe contingent effects of alliance managementcapability and new product development capabilitywith strong reference to the theoretical argumentdeveloped in the preceding section. In line with mymore general reasoning, I expect the relationshipbetween these two capabilities and competitiveadvantage to be the strongest when environmentaldynamism is at intermediate levels and compara-tively weaker when dynamism is low or high, aselaborated in greater detail below.

Alliance management capabilityand competitive advantage

The extant empirical literature finds that alliancemanagement capability tends to be positivelyrelated to performance (see Sluyts et al., 2011for a recent review). Organizations with a strongalliance management capability possess routinesthat support various alliance-related tasks, suchas partner identification and interorganizationallearning, that facilitate an effective execution of


184 O. Schilke

interfirm relationships (Schilke and Goerzen, 2010;Schreiner, Kale, and Corsten, 2009).

However, building and maintaining an alliancemanagement capability usually requires substantialinvestments in, for example, a dedicated alliancefunction that oversees and supports alliance oper-ations (Heimeriks and Duysters, 2007; Helfatet al., 2007; Kale, Dyer, and Singh, 2002). Sucha separate, specialized organizational unit cap-tures and codifies alliance-related knowledge fromongoing alliance relationships and disseminates itthroughout the firm. Other relevant investmentsmay include setting up alliance-specific intranetdatabases or holding regular alliance managementworkshops (Heimeriks, 2010).

While supporting the institutionalization ofalliance management capability, such investmentsare typically associated with nontrivial costs.Consistent with my general theoretical argumentregarding the amortization of dynamic capabilities,I suggest that such costs may not be fully justifiedwhen the firm has no need to employ alliancemanagement routines on a frequent basis—that it,when it only rarely engages in strategic alliances.One contextual factor that significantly affects theextent of alliance opportunities is environmentaldynamism. Analyzing alliance use of manufac-turing firms, Dickson and Weaver (1997) findthe dynamism of the environment to be a keydriver. Similarly, Rosenkopf and Schilling (2007)report that industries low in dynamism (such asclothing and construction supplies) also scored inthe lowest tertile for both alliance participationrates and number of alliances per firm. Thus, theextent to which firms engage in alliances depends(ceteris paribus) on the degree of environmen-tal dynamism, with low dynamism providingrelatively little need to make sufficient use ofalliance management routines so that the costsfrom alliance management capability would be faroutweighed by its gains. Beyond considerationsrelated to direct costs, another source of concernwhen investing in alliance management capabilityin relatively stable contexts with few needs foralliances is managers’ tendency to feel a necessityto legitimize those investments by promoting,and at times imposing, the use of alliancesand related management practices beyond afunctional level (Heimeriks, 2010). Based onthis reasoning, I suggest that the positive effectof alliance management capability in creating

competitive advantage is comparatively smallwhen environmental dynamism is low.

Further, consistent with my earlier generalargument regarding the effectiveness of dynamiccapabilities in highly dynamic environments, Ialso submit that very high levels of dynamismmay reduce the value-creation potential of alliancemanagement capability. This is because alliancemanagement capability rests on routinized prac-tices that leverage lessons learned from prioralliances (Anand and Khanna, 2000; Heimeriks,2010). In highly dynamic environments, how-ever, the nature of alliances may drastically dif-fer from one relationship to the next. Terjesen,Patel, and Covin (2011), for example, report sig-nificant positive associations between environmen-tal dynamism and alliance partner diversity aswell as alliance geographic diversity. Given thehigh degree of novelty that firms operating inhighly dynamic environments are likely to facein their alliances, matching appropriate routinesto these novel settings will prove challenging(matching problem). Additionally, highly dynamicenvironments may cause an inertia problem inthat alliance management capability may limita firm’s tendency to experiment with alternativebehavior. Continued reliance on established infor-mation transfer processes, for example, can pre-vent acquiring new types of knowledge that mayprove critical under drastically altered environ-mental conditions (Hoang and Rothaermel, 2005;Sampson, 2005). Also, firms with strong alliancemanagement capability tend to follow establishedpartner selection protocols (Heimeriks, 2010) andtend to engage in social bonding with their part-ners (Schreiner et al., 2009), both of which favorrepeated ties with the same portfolio of alliancepartners. Restricted partner selection, however,may prove particularly detrimental when operatingin highly dynamic environments where frequentlyswitching alliance partners is often required inorder to gain access to the currently most relevantresources (Kandemir, Yaprak, and Cavusgil, 2006).

At intermediate levels of dynamism, finally,I expect a balance to exist between firms’ abilityto leverage their alliance management investmentsand to effectively exploit their experience-basedalliance management routines. In these settings,alliances are frequent enough to justify the costsof developing alliance management capability, andenvironmental states are similar enough to pursuealliance management in a routinized fashion that



strongly builds on past experiences and to makeeffective use of similar types of alliances.

Hypothesis 1: The relationship between alliancemanagement capability and competitive advan-tage is strongest under intermediate levelsof environmental dynamism but comparativelyweaker when dynamism is low or high.

New product development capabilityand competitive advantage

A new product development capability is reflectedin organizational routines that structure innovationprocesses aimed at reconfiguring the firm’s productportfolio (Danneels, 2008; Lawson and Samson,2001; Subramaniam and Venkatraman, 2001). Itis commonly assumed that such routines leadto new product innovations that in turn resultin competitive advantage (Lawson and Samson,2001). However, there is reason to believe that thestrength of this positive effect varies across levelsof environmental dynamism.

Similar to my discussion of alliance manage-ment capability, it is important to note that a newproduct development capability usually entailsdurable commitment of funds—e.g., to supportskilled personnel, specialized facilities, and state-of-the-art equipment (Helfat et al., 2007). Forexample, Clark and Fujimoto (1990) find thatinvesting in specialized coordination committeespromotes routinized product development. Giventhe costs of such investments in developing newproduct development capability, firms need todeploy this capability repeatedly in order to gen-erate revenues from new or improved productsfor these expenses to pay off (Helfat and Winter,2011). Whereas new product launches and productoverhauls are critical to firms’ competitive advan-tage when contextual conditions change relativelyfrequently (Song et al., 2005), stable environmentsoften allow firms to sell existing products prof-itably without much alteration (Hambrick, 1983),making a new product development capability rel-atively less central to competitive advantage.

In highly dynamic environments, on the otherhand, product lifecycles tend to be compara-tively short and technological paradigm shiftsrelatively frequent. Although they provide ampleproduct development opportunities, I propose thatenvironments characterized by high dynamismpose considerable matching and inertia problems

that may decrease the relative effectiveness ofan experience-based new product developmentcapability. As Brown and Eisenhardt’s (1997)study illustrates, highly structured new productdevelopment processes are able to rapidly andflawlessly capture opportunities that build onprior product features, but these routines are oftenunable to accommodate opportunities that aredifferent in kind, suggesting a possible matchingproblem between unfamiliar environmental oppor-tunities and appropriate new product developmentactivities. Additionally, relying on experience-based new product development can result ininertia, which can prove particularly problematicwhen environmental change is frequent and dis-continuous. Firms with a strong established newproduct development capability tend to developa preference for pursuing incremental productimprovements along existing trajectories ratherthan exploring radically different innovations(Levinthal and Myatt, 1994; Sørensen and Stuart,2000). The empirical study by Leonard-Barton(1992) corroborates this view, showing that itwas precisely new product development routinesthat brought about dysfunctional restrictionsin exploring the scope of alternatives. Furtherillustrative evidence comes from Helfat et al.’s(2007: 49ff.) Rubbermaid case study. Long knownas a best-in-class ‘new product machine’ withhighly professionalized innovation routines thatallowed for continuously and quickly bringinga large number of products to the market, thefirm began to struggle when the environment wasbeginning to change drastically in the early 1990s.During that time, customers became significantlymore price conscious and large retailers such asWal-Mart gained substantial power. These werefundamental changes that Rubbermaid too longseemed to ignore while continuing to reinforceprevious recipes for new product innovationsuccess that no longer were appropriate, whichultimately resulted in a deterioration of the firm’scompetitive advantage.

Overall, I expect new product developmentcapability to be most valuable in moderately dy-namic contexts, where product innovation oppor-tunities occur in a relatively frequent but ratherincremental fashion. Extant qualitative compara-tive studies support the notion that environmentswith moderate dynamism provide an ideal contextfor new product development capability to unleashits greatest potential. In the moderately dynamic


186 O. Schilke

mainframe sector, for example, Eisenhardt andTabrizi (1995) find the capability’s underlyingroutines to substantially enhance predictabilityand effectiveness by coordinating the entirenew product development process from initialspecification through manufacturing ramp-upwhereas such routines were less beneficial in themore dynamic personal computing industry. Insummary, when environmental dynamism is at anintermediate level, there is a potential for repeatednew product launches that make investments incapability development worthwhile and firms alsohave the capacity to utilize effectively experience-based new product development routines to createnew, successful products that build on existingsolutions.

Hypothesis 2: The relationship between newproduct development capability and competitiveadvantage is strongest under intermediate lev-els of environmental dynamism but comparativelyweaker when dynamism is low or high .

DATA

The empirical research comprised three sequentialstages. I first conducted qualitative field interviewsto learn about types of capabilities relevant to orga-nizational resource reconfiguration, their potentialimplications for competitive advantage, as well asthe intelligibility of a preliminary survey question-naire. I next developed and conducted a large-scalesurvey. Three years later, I collected measures forthe dependent variable from the same firms thathad participated in the previous survey.

Qualitative field interviews

The fieldwork included 13 interviews with top-level managers from various industries. Each inter-view lasted between 45 and 90 minutes and con-sisted of three parts. In the first part, managerswere asked to elaborate on relevant types of rou-tine activities for adapting their organization tochanges in the environment. New product devel-opment and alliancing turned out to be among themost frequent responses.1 In the second part, I

1 Other routine activities that were mentioned pertained toinformation technology, marketing, and mergers.

scrutinized the study’s hypotheses by asking man-agers how critical these activities are for competi-tive advantage—both in general and, more specif-ically, when comparing environments character-ized by little, moderate, and substantial changes.There was considerable agreement that organiza-tional change routines can support firms’ com-petitive advantage. Managers disagreed, however,with regard to the relative performance impli-cations under varying degrees of environmentaldynamism. Mirroring the different perspectives inthe academic literature (see the literature review),some managers maintained that those routineswould be valuable in virtually any context. Otherssuggested that the strongest effect on competitiveadvantage should be observed in highly dynamicenvironments, whereas a few managers indicatedthat routine activities might prove comparativelyless useful in highly turbulent environments. Inthe third and final part of the interviews, managerswere asked to fill out a preliminary version of thequestionnaire to be used in the subsequent sur-vey study while providing feedback on the clarityof items as well as difficulties in responding tothem. As a result of this process, several question-naire items were reworded or eliminated. Anotherimportant insight came from a comment by twomanagers that, for diversified firms, all question-naire items should pertain to the business unitrather than the corporate level, as practices maydiffer substantially between business units, andmanagers can also provide more reliable infor-mation about the particular business unit they aremost strongly involved in.

Sample and data for the survey study

The two focal predictor variables in thehypotheses-testing survey study were alliancemanagement capability and new product devel-opment capability. In conceptualizing alliancemanagement capability, I followed prior allianceresearch (e.g., Eisenhardt and Schoonhoven,1996) and focused on alliances in research anddevelopment (R&D), given the diversity ofdifferent forms of alliances and their idiosyncraticgoals, policies, and structures. R&D alliances (asopposed to production or marketing alliances, forexample) have been argued to be more clearlydirected toward reconfiguring organizationalresources (Eisenhardt and Schoonhoven, 1996),



making them ideal for the purpose of studying aninstance of dynamic capability.

The study population comprised firms in thechemicals, machinery, and motor vehicle indus-tries: (1) because alliances are frequent in thesesectors (Hagedoorn, 1993); (2) because newproduct development activities play a key role inthese industries (Centre for European EconomicResearch, 2004); and (3) in order to capture a widevariance in the moderating variable environmentaldynamism. I obtained contact data for 2,226 firmsthrough Hoppenstedt Firmendatenbank, a largecommercial database containing a comprehensivelisting of firms located in Germany. Consistentwith the relationship criterion approach commonlyadopted in alliance research (Koka and Prescott,2002), I only included firms in this study that wereinvolved in at least one R&D alliance. For thispurpose, I employed a professional call center thatcontacted each of the 2,226 initial firms by tele-phone and determined whether they currently par-ticipated in R&D alliances.2 This led me to exclude840 firms that were not engaged in R&D alliances,resulting in a target population of 1,386 firms whowere asked for their participation in this study.

I received 302 usable responses, reflecting aresponse rate of 21.8 percent, which is consis-tent with comparable studies using key infor-mant methodology (e.g., Capron and Mitchell,2009). These 302 informants provided informa-tion on all constructs except for the dependentvariable (competitive advantage), which I mea-sured with a three-year time lag through a separatesurvey. My objectives were to establish tempo-ral order of the independent variables (precedingin time) to the dependent variables to enhancecausal inference (Biddle, Slavings, and Anderson,1985) as well as to allow time for the perfor-mance effects of dynamic capabilities to materi-alize (Zahra et al., 2006: 947) and also to reducethe threat of a potential common method bias thatcould have been present had I collected both inde-pendent and dependent variables simultaneously(Podsakoff and Organ, 1986). I chose a time lagof three years based on: (1) Rindfleisch et al.’s

2 The call center employees were trained extensively andprovided with a detailed interview guide. They were instructedto contact top-level managers, preferably heads of R&D ormembers of the executive board. Names of adequate contactpersons were partly extracted from Hoppenstedt or other publicsources and partly asked at the telephone switchboard.

(2008) assessment that three years is an appropri-ate compromise between enhancing causal infer-ence by implementing temporal order in the empir-ical design while not passing the outcome’s enddate; (2) Kor and Mahoney’s (2005: 495) find-ing in the medical instruments industry that ‘R&Dinvestments convert into revenue-generating prod-ucts typically within a period of three years’; and(3) prior usage of a three-year lag in longitudinalsurvey studies on strategic alliances (Rindfleischand Moorman, 2001, 2003). Of the 302 firms thatresponded to the first survey in 2006, nine hadceased to exist because they were acquired or dis-solved. In the remaining 293 firms, I contacted thesame key informant who had participated threeyears earlier. After several reminders, I received204 responses from these informants. In order tofurther increase the number of responses, I triedto contact an alternative top manager if the origi-nal informant was no longer available or remainedunresponsive. This allowed me to gather informa-tion on competitive advantage from an additional75 firms; thus, the study’s final sample consistsof 279 matched questionnaires across times 1 and2. While this sample size may not be consid-ered very large, it is much in line with samplesizes in other strategy studies (Phelan, Ferreira,and Salvador, 2002) and exceeds common recom-mendations for advanced statistical analyses (e.g.,MacCallum et al., 1999).

Characteristics of the firms and informants inthe sample are provided in Table 1. To verify theappropriateness of the key informants, question-naire items asked about their tenure and expertise(Kumar, Stern, and Anderson, 1993). Overall,73.5 percent of the participants in the final datasethad been with their current firm for six yearsor longer (see Table 1). In addition, I assessedrespondents’ self-reported knowledge of the firm’sR&D alliances and innovation-related activities onfive-point answer scales ranging from 1 (‘poor’)to 5 (‘excellent’). The means of 4.07 (SD = 0.84)and 4.12 (SD = 0.72), respectively, suggested thatthe informants were very well informed.

I checked for nonresponse bias in three ways.First, I assessed a nonresponse bias by compar-ing early and late respondents (Armstrong andOverton, 1977). The results of the t-tests indicatedno significant differences (p > 0.05) across meansfor each of the theoretical constructs betweenearly and late respondents. Second, I examinedwhether the nonresponding firms differed from the


188 O. Schilke

Table 1. Sample composition

Sample int = 1

(n = 302)(%)

Sample int = 2

(n = 279)(%)

IndustryMachinery 55.2 54.1Chemicals 21.0 22.6Motor vehicles 23.8 23.3

Firm size 0.05). These findings pro-vide consistent evidence that nonresponse bias isnot a problem. Kruskal Wallis H tests also showedno significant differences in responses of the fourinformant groups (i.e., heads of R&D, project lead-ers in R&D, members of the executive board, mis-cellaneous).

Measures

I used multi-item scales to measure the inde-pendent, dependent, and moderating variables.Consistent with the qualitative interviews, if therespondent worked for a diversified firm, he/shewas asked to answer all questions with referenceto the business unit for which he/she worked.3

Table 2 lists the measurement items used to opera-tionalize the constructs. When adequate measureswere available, I adapted them from prior stud-ies. Following the recommendations of DeVel-lis (2003), the questionnaire items were furtherrefined through in-depth interviews with 13 man-agers (described above), an item sorting pretestbased on Anderson and Gerbing (1991) adminis-tered to 15 scholars, and a pretest of the question-naire conducted with 21 managers. When possible,survey information obtained from the key infor-mant in the main study was triangulated with com-plementary data sources to establish its accuracy(Homburg et al., 2012), as described below.

Competitive advantage

A firm is said to have a competitive advan-tage when it enjoys greater success than currentor potential competitors in its industry, suggest-ing that superior firm performance serves as akey indicator of competitive advantage (Barnett,Greve, and Park, 1994; Ghemawat and Rivkin,1999). Specifically, I operationalized competitiveadvantage as a two dimensional construct, withthe first-order dimensions of (1) strategic per-formance (qualitative dimension) and (2) finan-cial performance (quantitative dimension), both ofwhich were measured in comparison to competi-tion. Items for the two performance dimensionswere adapted from Jap (1999) and Weerawardena(2003).

To corroborate the performance informationobtained from key informants, I collected account-ing performance data for a subset of 48 companiesfor which such information was available. Usinga public financial database and company reportsavailable on the firms’ websites, I obtained infor-mation on return on investment (ROI) and return

3 While the fact that the sample consists of both firms andbusiness units may be viewed as a limitation, I control forthis issue in the empirical analysis as described further below.Reported results are also robust to dropping business units.



Tabl

e2.

Mea

sure

men

tsc

ales

Mea

n/SD

Com

peti

tive

adva

ntag

e4.

80/0

.95

(‘St

rong

lydi

sagr

ee’

[1]

to‘s

tron

gly

agre

e’[7

])St

rate

gic

perf

orm

ance

α=

0.73

CR

=0.

75A

VE

=0.

501a

We

have

gain

edst

rate

gic

adva

ntag

esov

erou

rco

mpe

titor

s.5.

29/1

.10

1bW

eha

vea

larg

em

arke

tsh

are.

5.13

/1.3

91c

Ove

rall,

we

are

mor

esu

cces

sful

than

our

maj

orco

mpe

titor

s.4.

98/1

.20

Fin

anci

alpe

rfor

man

ceα

=0.

93C

R=

0.93

AV

E=

0.81

2aO

urE

BIT

(ear

ning

sbe

fore

inte

rest

and

taxe

s)is

cont

inuo

usly

abov

ein

dust

ryav

erag

e.4.

49/1

.37

2bO

urR

OI

(ret

urn

onin

vest

men

t)is

cont

inuo

usly

abov

ein

dust

ryav

erag

e.4.

42/1

.27

2cO

urR

OS

(ret

urn

onsa

les)

isco

ntin

uous

lyab

ove

indu

stry

aver

age.

4.47

/1.3

1A

llian

cem

anag

emen

tca

pabi

lity

4.61

/1.0

6(‘

Stro

ngly

disa

gree

’[1

]to

‘str

ongl

yag

ree’

[7])

Inte

rorg

aniz

atio

nal

coor

dina

tion

α=

0.85

CR

=0.

85A

VE

=0.

593a

Our

activ

ities

with

R&

Dal

lianc

epa

rtne

rsar

ew

ell

coor

dina

ted.

5.18

/1.6

73b

We

ensu

reth

atou

rw

ork

task

sfit

with

thos

eof

our

R&

Dal

lianc

epa

rtne

rsve

ryw

ell.

4.84

/1.7

03c

We

ensu

reth

atou

rw

ork

issy

nchr

oniz

edw

ithth

ew

ork

ofou

rR

&D

allia

nce

part

ners

.4.

59/1

.67

3dT

here

isa

grea

tde

alof

inte

ract

ion

with

our

R&

Dal

lianc

epa

rtne

rson

mos

tde

cisi

ons.

4.96

/1.5

3A

llian

cepo

rtfo

lioco

ordi

nati

onα

=0.

91C

R=

0.91

AV

E=

0.73

4aW

een

sure

anap

prop

riat

eco

ordi

natio

nam

ong

the

activ

ities

ofou

rdi

ffer

ent

R&

Dal

lianc

es.

4.25

/1.7

2

4bW

ede

term

ine

area

sof

syne

rgy

inou

rR

&D

allia

nce

port

folio

.4.

32/1

.74

4cW

een

sure

that

inte

rdep

ende

ncie

sbe

twee

nou

rR

&D

allia

nces

are

iden

tified

.4.

02/1

.67

4dW

ede

term

ine

ifth

ere

are

over

laps

betw

een

our

diff

eren

tR

&D

allia

nces

.4.

59/1

.73

Inte

rorg

aniz

atio

nal

lear

ning

α=

0.88

CR

=0.

88A

VE

=0.

645a

We

have

the

capa

bilit

yto

lear

nfr

omou

rR

&D

allia

nce

part

ners

.5.

20/1

.37

5bW

eha

veth

em

anag

eria

lco

mpe

tenc

eto

abso

rbne

wkn

owle

dge

from

our

R&

Dal

lianc

epa

rtne

rs.

4.71

/1.4

7

5cW

eha

vead

equa

tero

utin

esto

anal

yze

the

info

rmat

ion

obta

ined

from

our

R&

Dal

lianc

epa

rtne

rs.

4.93

/1.4

9

5dW

eca

nsu

cces

sful

lyin

tegr

ate

our

exis

ting

know

ledg

ew

ithne

win

form

atio

nac

quir

edfr

omou

rR

&D

allia

nce

part

ners

.4.

86/1

.48

Alli

ance

proa

ctiv

enes

sα

=0.

88C

R=

0.88

AV

E=

0.65

6aW

est

rive

topr

eem

ptou

rco

mpe

titio

nby

ente

ring

into

R&

Dal

lianc

eop

port

uniti

es.

4.46

/1.7

46b

We

ofte

nta

keth

ein

itiat

ive

inap

proa

chin

gfir

ms

with

R&

Dal

lianc

epr

opos

als.

4.20

/1.7

16c

Com

pare

dto

our

com

petit

ors,

we

are

far

mor

epr

oact

ive

and

resp

onsi

vein

findi

ngan

d‘g

oing

afte

r’R

&D

part

ners

hips

.3.

89/1

.61

6dW

eac

tivel

ym

onito

rou

ren

viro

nmen

tto

iden

tify

R&

Dpa

rtne

rshi

pop

port

uniti

es.

4.57

/1.5

0


190 O. Schilke

Tabl

e2.

(Con

tinu

ed)

Mea

n/SD

Alli

ance

tran

sfor

mat

ion

α=

0.83

CR

=0.

83A

VE

=0.

627a

We

are

will

ing

topu

tas

ide

cont

ract

ual

term

sto

impr

ove

the

outc

ome

ofou

rR

&D

allia

nces

.4.

18/1

.60

7bW

hen

anun

expe

cted

situ

atio

nar

ises

,w

ew

ould

rath

erm

odif

yan

R&

Dal

lianc

eag

reem

ent

than

insi

ston

the

orig

inal

term

s.4.

90/1

.46

7cFl

exib

ility

,in

resp

onse

toa

requ

est

for

chan

ge,

isch

arac

teri

stic

ofou

rR

&D

allia

nce

man

agem

ent

proc

ess.

4.89

/1.4

3

New

prod

uct

deve

lopm

ent

capa

bilit

yα

=0.

81C

R=

0.83

AV

E=

0.55

Obj

ectiv

esfo

run

dert

akin

gin

nova

tion

proj

ects

inth

ela

stth

ree

year

s:(‘

Not

impo

rtan

t’[1

]to

‘ver

yim

port

ant’

[7])

8aIn

trod

uce

new

gene

ratio

nof

prod

ucts

5.77

/1.3

48b

Ext

end

prod

uct

rang

e5.

42/1

.42

8cO

pen

upne

wm

arke

ts.

6.02

/1.0

48d

Ent

erne

wte

chno

logy

field

s4.

68/1

.58

Env

iron

men

tal

dyna

mis

mα

=0.

81C

R=

0.81

AV

E=

0.47

(‘St

rong

lydi

sagr

ee’

[1]

to‘s

tron

gly

agre

e’[7

])9a

The

mod

esof

prod

uctio

n/se

rvic

ech

ange

ofte

nan

din

am

ajor

way

.2.

73/1

.18

9bT

heen

viro

nmen

tal

dem

ands

onus

are

cons

tant

lych

angi

ng.

3.80

/1.4

79c

Mar

ketin

gpr

actic

esin

our

indu

stry

are

cons

tant

lych

angi

ng.

3.18

/1.2

79d

Env

iron

men

tal

chan

ges

inou

rin

dust

ryar

eun

pred

icta

ble.

3.51

/1.4

09e

Inou

ren

viro

nmen

t,ne

wbu

sine

ssm

odel

sev

olve

freq

uent

ly.

2.85

/1.2

9A

llian

cepo

rtfo

liosi

ze10

aH

owm

any

allia

nces

isyo

urfir

m/b

usin

ess

unit

invo

lved

inat

pres

ent?

7.08

/18.

68P

rodu

ctsc

ope

Stra

tegi

cob

ject

ives

:(‘

Not

impo

rtan

t’[1

]to

‘ver

yim

port

ant’

[7])

11a

Bre

adth

ofpr

oduc

tof

feri

ng(p

ursu

ing

ana

rrow

,fo

cuse

dpr

oduc

tsc

ope)

4.67

/1.4

5M

arke

tsc

ope

Stra

tegi

cob

ject

ives

:(‘

Not

impo

rtan

t’[1

]to

‘ver

yim

port

ant’

[7])

12a

Bre

adth

ofta

rget

edm

arke

tse

gmen

ts(p

ursu

ing

ana

rrow

,fo

cuse

dm

arke

tsc

ope)

3.85

/1.5

2P

roce

ssin

nova

tion

(‘St

rong

lydi

sagr

ee’

[1]

to‘s

tron

gly

agre

e’[7

])13

aW

eha

vefr

eque

ntly

impr

oved

man

ufac

turi

ngor

oper

atio

nal

proc

esse

s.5.

04/1

.45

Fir

mun

itof

anal

ysis

(‘N

o’[0

],‘y

es’

[1])

14a

Plea

sein

dica

tew

heth

eryo

urre

spon

ses

pert

ain

toa

non-

dive

rsifi

edfir

m(a

sop

pose

dto

abu

sine

ssun

itin

adi

vers

ified

firm

with

seve

ral

busi

ness

units

).0.

89/0

.32



on sales (ROS) for each of the three years preced-ing the second survey. I then computed the averageROI and ROS for those years and standardized themeasures by industry. Subsequently, I correlatedthis archival data with perceptual responses aver-aged across the six competitive advantage items.Both measures were significantly correlated (ROI:r = 0.44, p ≤ 0.001; ROS: r = 0.48, p ≤ 0.001).Although these correlations were relatively lowercompared to what Robson, Katsikeas, and Bello(2008) obtained using a similar approach, theycompare favorably to several other studies report-ing correlations between subjective and archivalperformance data (e.g., Boyer, 1999; Douglas andJudge, 2001; Krishnan, Martin, and Noorderhaven,2006).

To provide further evidence for sufficient accu-racy, I gathered performance information froma second key informant in a total of 36 firmsand calculated ICC(1) to determine the level ofinterrater reliability. I obtained an ICC(1) of 0.24,which clearly exceeded Bliese’s (1998) 0.1 cutoff.Finally, I relied on information on organizationalgrowth, which population ecologists often use asa proxy for competitive advantage (Baum, 1996),to triangulate the dependent variable. For thefirms for which such information was available,I computed three-year percentage changes insales revenues (n = 48), number of employees(n = 279), and accounting value of assets (n = 48)(Helfat et al., 2007) and then correlated thesethree measures with the average of the itemsof the competitive advantage construct. I foundsignificant associations for growth in sales rev-enues (r = 0.32, p ≤ 0.01), number of employees(r = 0.28, p ≤ 0.001), and accounting value ofassets (r = 0.39, p ≤ 0.001), which lends furthercredibility to the perceptual competitive advantagemeasure.

Alliance management capability

I used the measure developed by Schilke andGoerzen (2010), which suggests a five-dimen-sional, second-order structure of the construct,with the underlying dimensions of (1) interor-ganizational coordination; (2) alliance portfoliocoordination; (3) interorganizational learning; (4)alliance proactiveness; and (5) alliance transfor-mation. Interorganizational coordination pertainsto the governance of individual alliances, whereas

alliance portfolio management involves the inte-gration of the firm’s various strategic alliances.Interorganizational learning reflects routinesdesigned to facilitate knowledge transfers acrossorganizational boundaries. Alliance proactivenesscan be defined as routine efforts to identify poten-tially valuable partnering opportunities. Finally,alliance transformation concerns routines to mod-ify alliances over the course of the alliance process.

I corroborated the subjective alliance manage-ment capability measure by correlating it withthe firm’s prior alliance experience, a widelyused proxy for alliance management capability(Anand and Khanna, 2000; Hoang and Rothaer-mel, 2005). To measure alliance experience,I asked respondents to indicate the number ofprior agreements with R&D alliance partnerswithin the last five years and used a logarithmictransformation to correct skewness. This variablewas then correlated with the composite scoreof the alliance management capability construct,computed as the simple average of its dimensions’items. Both measures were significantly correlated(r = 0.27; p ≤ 0.001), which supported the validityof the perceptual measure.

New product development capability

To capture the firm’s new product developmentcapability, I relied on the measurement itemsintroduced by He and Wong (2004). These itemsgauge the extent to which a firm routinely carriesout innovation projects aimed at entering newproduct domains. I triangulated this measure witharchival information on R&D intensity (R&Dexpenditures divided by revenues), which hasoften been used as a proxy for innovation-relateddynamic capabilities in archival research (Helfat,1994a, 1994b, 1997). For the 48 firms for whichrelevant secondary data were available, I found astrong positive association with the average of thesurvey items (r = 0.30; p ≤ 0.001).

Environmental dynamism

Environmental dynamism refers to the volatilityand unpredictability of the firm’s external envi-ronment (Miller and Friesen, 1983). To capturedynamism, I used items developed by Miller andFriesen (1982) and Jap (1999). For the purposeof validating managers’ perceptions of environ-mental dynamism, I applied two archival indexes


192 O. Schilke

measuring instability in sales and net assets(Sutcliffe, 1994). To compute these indexes, Iregressed sales and net assets for a period of threeyears prior to the survey on a variable representingthe time period and divided the standard errors ofthe regression by the mean level of the dependentvariable (Dess and Beard, 1984). Correlationsof these indexes with the subjective measure ofdynamism were 0.36 (n = 48) and 0.38 (n = 48),respectively; both were significant at p ≤ 0.001.These positive and highly significant correlationsexceeded those obtained by Sharfman and Dean(1991) in a similar analysis and supported thevalidity of the perceptual measure of environ-mental dynamism. Furthermore, complementaryperceptual information from 36 secondary keyinformants was used to determine interraterreliability. I obtained an ICC(1) of 0.20, whichclearly exceeds the common 0.1 threshold.

Control variables

Consistent with Li, Poppo, and Zhou (2008), I con-sidered industry effects, firm size, and firm age ascontrols. In addition, I controlled for the firm’salliance portfolio size, product scope, marketscope, and process innovation, responses pertain-ing to either a firm or a business unit, and the use ofeither the same or a different respondent during thesecond data collection wave, as elaborated below.

1. Industry effects . The importance of the industryin which a firm competes as a predictor offirm-level variables is widely recognized inthe literature (Dess, Ireland, and Hitt, 1990).To control for industry effects, I used dummyvariables, specifying the chemicals industry asthe base to which the effects of the otherdummies (machinery and motor vehicles) werecompared.

2. Firm age. Firm age has been suggested to influ-ence a firm’s competitive advantage (Zahra,Ireland, and Hitt, 2000) as well as the extentof patterned forms of behavior that underpindynamic capabilities (Helfat and Peteraf, 2003).I measured firm age in terms of the number ofyears since the establishment of the firm, clas-sifying the number of years into six categories(ranging from 1 for firms that are younger than5 years to 6 for firms that are 50 years or older)(Capron and Mitchell, 2009).

3. Firm size. Firm size can enhance competitiveadvantage by, for example, facilitating accessto a lower cost of capital while simultaneouslylowering risk (Chang and Thomas, 1989). Firmsize may also influence the firm’s dynamiccapabilities, with larger firms being able todedicate more resources to developing theirchange routines. Size was assessed based ona firm’s total number of full-time employees(ranging from 1 for firms that have fewer than100 employees to 6 for firms that have 5,000or more employees).

4. Alliance portfolio size. Previous research hasassociated a firm’s number of alliances withperformance outcomes (Powell, Koput, andSmith-Doerr, 1996) and with innovation inten-sity (Hagedoorn and Schakenraad, 1994). Addi-tionally, firms with a large alliance portfoliocan be expected to have strongly institutional-ized alliance management routines. I measuredalliance portfolio size by the firm’s total numberof current alliances (Jiang, Tao, and Santoro,2010) and logarithmized this measure to reduceskewness.

5. Product and market scope. In line with Zott andAmit (2008), I controlled for the breadth of thefirm’s product offering and targeted market, asthese are key dimensions of a firm’s strategythat may affect its competitive advantage andcapability development. I adapted the question-naire items for these two variables from Zottand Amit (2008).

6. Process innovation . Process innovation refersto the introduction of new elements into anorganization’s operations. I measured processinnovation with the item ‘We have frequentlyimproved manufacturing or operational pro-cesses,’ which has previously been used by Su,Tsang, and Peng (2009).

7. Firm unit of analysis . Because the sample com-prises one set of observations for firms andanother set of observations for business unitswithin firms (as mentioned above), I followedthe approach by Mithas, Ramasubbu, and Sam-bamurthy (2011) and used a dummy (1 = firmsand 0 = business units) to account for this dif-ference.

8. Same respondent . I used a dummy variable tocontrol for the fact that in a subset of firms,the informant used in t = 2 differed from theinformant used in t = 1. The dummy was coded



as 1 when the identical respondent was used inboth waves of data collection.

Measurement properties of constructs

Table 2 reports coefficient alphas (α), compositereliabilities (CR), and average variances extracted(AVE) for the study’s first-order, multi-item con-structs. The values obtained indicate reliable andvalid measures of the individual constructs. Afterassessing the constructs individually, I performeda confirmatory factor analysis among all first-order factors, using the structural equation mod-eling software AMOS 16.0 (Arbuckle, 2007) andthe maximum likelihood (ML) procedure (Hairet al., 2006). The measures of goodness of fithad satisfactory values (χ2 = 1,013.80; df = 741;χ2/df = 1.37; CFI = 0.95; GFI = 0.87; TLI = 0.94;RMSEA = 0.04). Following Fornell and Larcker(1981), I assessed the discriminant validity of thefactors in the model and found that the square rootof the average variance extracted by the measure ofeach factor is larger than the absolute value of thecorrelation of that factor’s measure with all mea-sures of other factors in the model, as reported inTable 3.

Common method bias

Although using key informants is common inresearch on organizational capabilities in orderto obtain required data on intrafirm processes(e.g., Capron and Mitchell, 2009; Danneels, 2008;Gruber et al., 2010; Kemper, Schilke, and Bret-tel, forthcoming), common method bias mightpose a problem in such studies (Podsakoff andOrgan, 1986). To safeguard against this possi-bility, I undertook several steps. First, and mostimportantly, measures of the dependent variablewere collected in a separate survey (Podsakoffand Organ, 1986). Second, I performed Harman’sone-factor test by loading all indicators of thestudy constructs into an exploratory factor analy-sis. Results revealed that no single factor explainedmore than 30 percent of the total variance inthe variables, suggesting that common methodbias was unlikely to be a serious problem inthis study. Additionally, I also applied Harman’sone-factor test using confirmatory factor analyses(McFarlin and Sweeney, 1992), which compareda single-factor model with the proposed 19-factormodel. Results showed that the single-factor model

had a significantly worse fit (χ2diff = 1,232.33;dfdiff = 170; p ≤ 0.01). These findings, along withthose reported earlier regarding the significantassociations between subjective and archival mea-sures, indicated that common method bias was nota serious concern in this study.

METHOD AND RESULTS

To test the hypotheses, I analyzed nonlinear inter-actions using OLS regression based on the pro-cedure outlined by Jaccard (2003). This involvedaveraging the items for each construct (in caseof a multi-dimensional construct, averaging theitems for all of the construct’s dimensions), mean-centering interacting variables, calculating thesquare of the moderating variable (environmentaldynamism), constructing linear as well as squaredproduct terms, and finally estimating the followingregression equation:

Competitive advantage = a + b1 machinery+ b2 motor vehicles + b3 firm age+ b4 firm size + b5 alliance portfolio size+ b6 product scope + b7 market scope+ b8 process innovation+ b9 firm unit of analysis+ b10 same respondent+ b11 alliance management capability+ b12 new product development capability+ b13 environmental dynamism+ b14 environmental dynamism squared+ b15 alliance management capability× environmental dynamism+ b16 new product development capability× environmental dynamism+ b17 alliance management capability× environmental dynamism squared+ b18 new product development capability× environmental dynamism squared + e

A significant coefficient of the squared moderatorproduct term (here: b17 and b18) would indicate the


194 O. Schilke

Tabl

e3.

Des

crip

tive

stat

istic

san

ddi

scri

min

ant

valid

ity

Fact

orSc

ale

rang

eM

ean

SD1

23

45

67

89

1011

1213

1415

1617

1819

1St

rate

gic

perf

orm

ance

1–

75.

130.

990.

712

Fina

ncia

lpe

rfor

man

ce1

–7

4.46

1.23

0.56

0.90

3In

tero

rgan

izat

iona

lco

ordi

natio

n1

–7

4.89

1.37

0.33

0.16

0.77

4A

llian

cepo

rtfo

lioco

ordi

natio

n1

–7

4.30

1.52

0.28

0.25

0.63

0.85

5In

tero

rgan

izat

iona

lle

arni

ng1

–7

4.93

1.24

0.33

0.29

0.68

0.60

0.80

6A

llian

cepr

oact

iven

ess

1–

74.

281.

410.

440.

330.

610.

540.

670.

817

Alli

ance

tran

sfor

mat

ion

1–

74.

661.

290.

160.

120.

500.

340.

510.

490.

798

New

prod

uct

deve

lopm

ent

capa

bilit

y1

–7

5.47

1.09

0.56

0.30

0.39

0.29

0.43

0.49

0.25

0.74

9E

nvir

onm

enta

ldy

nam

ism

1–

73.

211.

00−0

.03

0.07

−0.0

20.

03−0

.04

0.04

0.01

0.11

0.69

10M

achi

nery

0/1

0.54

0.50

0.09

−0.0

6−0

.09

−0.0

6−0

.12

−0.1

0−0

.05

0.00

0.03

n/a

11M

otor

vehi

cles

0/1

0.23

0.42

−0.0

4−0

.05

0.12

0.03

0.08

0.03

0.03

−0.0

2−0

.07

−0.6

0n/

a12

Firm

age

1–

64.

991.

380.

170.

030.

040.

000.

060.

030.

160.

09−0

.08

0.01

−0.0

2n/

a13

Firm

size

1–

63.

251.

370.

240.

160.

200.

150.

190.

170.

030.

19−0

.08

−0.2

20.

260.

28n/

a14

Alli

ance

port

folio

size

0-∞

1.32

0.92

0.15

0.08

0.11

0.13

0.18

0.28

0.01

0.24

0.08

−0.0

6−0

.11

0.02

0.22

n/a

15Pr

oduc

tsc

ope

1–

74.

671.

45−0

.01

−0.0

4−0

.01

−0.0

1−0

.04

0.00

0.15

−0.0

40.

220.

15−0

.05

−0.0

4−0

.16

−0.0

7n/

a16

Mar

ket

scop

e1

–7

3.85

1.52

−0.0

5−0

.12

−0.0

3−0

.04

0.02

−0.0

5−0

.03

−0.0

10.

100.

040.

03−0

.03

−0.2

1−0

.10

0.42

n/a

17Pr

oces

sin

nova

tion

1–

75.

041.

450.

170.

090.

260.

130.

180.

270.

170.

470.

160.

12−0

.05

−0.0

90.

000.

12−0

.03

−0.0

2n/

a18

Firm

unit

ofan

alys

is0/

10.

890.

32−0

.03

−0.0

2−0

.09

−0.1

4−0

.09

0.00

−0.0

2−0

.05

0.02

0.05

−0.0

7−0

.04

−0.2

2−0

.09

0.00

−0.0

10.

10n/

a19

Sam

ere

spon

dent

0/1

0.73

0.44

−0.1

0−0

.13

0.00

−0.0

1−0

.03

0.00

−0.0

5−0

.08

−0.0

40.

07−0

.05

−0.0

3−0

.11

0.07

0.03

0.03

0.01

0.01

n/a

n=

279;

num

bers

onth

edi

agon

alsh

owsq

uare

root

sof

AV

E,

num

bers

belo

wth

edi

agon

alsh

owco

rrel

atio

ns;

AV

Eno

tav

aila

ble

for

sing

le-i

tem

cons

truc

ts;

corr

elat

ions

with

abso

lute

valu

e>

0.17

are

sign

ifica

ntat

the

1%le

vel

and

>0.

12at

the

5%le

vel.



presence of quadratic moderation, suggesting thatthe relationship between the independent variableand the outcome varies as a nonlinear functionof the moderator. More specifically, a positivecoefficient suggests a U-shaped pattern whereas anegative coefficient indicates an inverse U-shapedpattern, the latter of which would be in line withthe hypotheses.

Table 4 summarizes the regression results.Model 1 includes controls only, and model 2adds the direct effects of alliance managementcapability, new product development capability,and environmental dynamism. Model 3 addition-ally includes linear interaction terms. Model 4 ismy main model and introduces squared interac-tion terms. Inspection of variance inflation factors(VIF) among the explanatory variables in all fourmodels revealed the highest VIF to be 2.49. Thissuggests that no problematic multicollinearity ispresent (Kleinbaum, Kupper, and Muller, 1988).Regarding the main effects in model 4, the regres-sion coefficient of 0.37 indicates a positive andhighly significant (p ≤ 0.01) relationship betweenalliance management capability and competitiveadvantage. The coefficient of new product devel-opment capability shows that firms with a strongernew product development capability have a sig-nificantly higher competitive advantage (b = 0.39;p ≤ 0.01). As such, both dynamic capabilities havea positive relationship with competitive advantage.Among the control variables, firm size is signifi-cantly related to competitive advantage (b = 0.12;p ≤ 0.01).

With regard to the hypotheses, the negative andhighly significant coefficients of the two squaredproduct terms suggest that the relationshipsbetween the two dynamic capabilities and com-petitive advantage vary across different levels ofenvironmental dynamism in a quadratic manner.The nature of the interactions is illustrated inFigure 1. The graphs in this figure represent asso-ciations between alliance management capabilityand competitive advantage (Figure 1a) and newproduct development capability and competitiveadvantage (Figure 1b) across different levels ofenvironmental dynamism. To create these graphs,the regression equation was examined at differentlevels of environmental dynamism, using themargins command implemented in STATA 11.The vertical axes of the graphs represent valuesof regression coefficients for alliance manage-ment capability and new product development

capability, respectively; and the horizontal axesrepresent values of environmental dynamismbetween two standard deviations below and abovethe mean (i.e., between 1.21 and 5.21).

The proposed inverse U-shaped relationshipbetween dynamic capability and competitiveadvantage across increasing levels of environ-mental dynamism is apparent in both graphs. Asshown in Figure 1(a), for firms that experience alow or a high level of environmental dynamism,the coefficient for the regression of competitiveadvantage on alliance management capabilityis comparatively low and, at very low levelsof environmental dynamism, nonsignificant.However, at intermediate levels of environmentaldynamism, the association was strongly positiveand significant. Figure 1(b) shows an analogousinverse U-shaped graph for the regression of com-petitive advantage on new product developmentcapability. These illustrations, together with thesignificant quadratic interaction terms, provideempirical support for Hypotheses 1 and 2.

In comparing the two graphs, it becomes appar-ent that the range for which the respective regres-sion coefficient is significant is located at slightlyhigher levels of environmental dynamism in thecase of alliance management capability (between2.3 and 4.9) as compared to new product develop-ment capability (between 2.1 and 4.4). This obser-vation suggests that alliance management capabil-ity has a positive impact on competitive advan-tage at relatively higher levels of environmentaldynamism when compared to new product devel-opment capability.

POST-HOC ANALYSES

Four supplemental analyses demonstrated the ro-bustness of the results. First, I conducted the Haus-man (1978) endogeneity test (e.g., Wooldridge,2008), using two instruments that have previouslybeen identified as correlates of dynamic capabili-ties: willingness to cannibalize and organizationalslack (Danneels, 2008). The first instrument wasmeasured with the item ‘We support projects evenif they could potentially take away sales fromexisting products’ and the second instrument wascaptured by ‘My firm has a reasonable amount ofresources in reserve’ (Danneels, 2008). Hausman’s(1978) endogeneity test was not significant foreither alliance management capability or new


196 O. Schilke

Table 4. Regression results

Variables Coefficient Model 1 Model 2 Model 3 Model 4

Intercept a 3.94** 2.33** 4.84** 4.91**

(0.44) (0.46) (0.42) (0.39)Controls

Machinery b1 −0.05 0.01 0.00 0.05(0.14) (0.13) (0.13) (0.12)

Motor vehicles b2 −0.24 −0.22 −0.23 −0.09(0.17) (0.16) (0.16) (0.15)

Firm age b3 0.03 0.00 0.00 0.00(0.04) (0.04) (0.04) (0.04)

Firm size b4 0.14** 0.10** 0.11** 0.12**

(0.05) (0.04) (0.04) (0.04)Alliance portfolio size b5 0.05 −0.03 −0.06 −0.09

(0.06) (0.06) (0.06) (0.06)Product scope b6 0.02 0.01 0.01 0.02

(0.04) (0.04) (0.04) (0.04)Market scope b7 −0.03 −0.04 −0.04 −0.05

(0.04) (0.04) (0.04) (0.04)Process innovation b8 0.09** −0.05 −0.04 −0.04

(0.04) (0.04) (0.04) (0.04)Firm unit of analysis b9 0.02 0.12 0.16 0.10

(0.18) (0.16) (0.16) (0.15)Same respondent b10 −0.24* −0.20† −0.21† −0.16

(0.13) (0.11) (0.11) (0.11)Predictors

Alliance management capability b11 0.22** 0.22** 0.37**

(0.05) (0.05) (0.07)New product development

capabilityb12 0.27** 0.28** 0.39**

(0.06) (0.06) (0.07)Environmental dynamism b13 0.03 0.01 0.05

(0.05) (0.05) (0.05)Environmental dynamism squared b14 0.00

(0.04)Alliance management

capability × environmentaldynamism

b15 0.08 0.18**

(0.05) (0.06)New product development

capability × environmentaldynamism

b16 0.11† 0.04

(0.06) (0.05)Alliance management

capability × environmentaldynamism squared

b17 −0.16**

(0.04)New product development

capability × environmentaldynamism squared

b18 −0.17**

(0.05)R-squared 0.10 0.26 0.28 0.37Adjusted R-squared 0.06 0.22 0.24 0.33

n = 279; unstandardized coefficients and standard errors (in parentheses) are reported.†p ≤ 0.10; *p ≤ 0.05; **p ≤ 0.01.



(a) (b)

Figure 1. The relationship between dynamic capabilities and competitive advantage as a function of environmentaldynamism (with 95% confidence interval). (a) Alliance management capability. (b) New product development capability

product development capability (p > 0.1), whichattenuated concerns of endogeneity in the empir-ical analysis. Second, I reestimated the regressionmodel using strategic and financial performance(instead of the competitive advantage construct)as dependent variables. Results did not changequalitatively from the original model specification.The effects of alliance management capability andnew product development capability remainedpositive and statistically significant at p ≤ 0.01,and the effects of the squared interaction termsremained negative and statistically significant atp ≤ 0.01 in both alternative models. Third, asan alternative approach for examining nonlinearmoderation, I estimated a spline (instead ofa polynomial) specification, in which I brokeenvironmental dynamism into linear splinesknotted at the median and interacted these splineswith alliance management capability and newproduct development capability. The interactionswere positive up to the median and then becamenegative, in line with Hypotheses 1 and 2 (fullresults for this specification are available uponrequest). Fourth, I also used multi-group structuralequation modeling to test for moderation (Byrne,2001; Hair et al., 2006). Please see the AppendixS1 for details. The results of the multi-groupanalyses lent further support to the hypotheses.

DISCUSSION

This paper presented two hypotheses suggest-ing that the effects of both alliance management

capability and new product development capabilityon a firm’s competitive advantage vary as a non-linear function of environmental dynamism. Morespecifically, building on dynamic capabilities the-ory as well as alliance and new product develop-ment literature, I proposed that these two capabili-ties would have the strongest positive impact oncompetitive advantage under intermediate levelsof environmental dynamism, whereas their impactwould be comparatively weaker in stable andhighly dynamic contexts. I tested these hypothe-ses empirically and found strong support for myposition. The analyses indicated that the effects ofthe two capabilities on competitive advantage arehighest when environmental dynamism is moder-ate and comparatively lower when environmentaldynamism is low or high.

Two somewhat contradictory positions exist onthe value of dynamic capabilities under differ-ent levels of environmental dynamism. One sug-gests that their effect on competitive advantage iscomparatively smaller at low levels of dynamism(Drnevich and Kriauciunas, 2011; Helfat et al.,2007; Winter, 2003; Zahra et al., 2006; Zolloand Winter, 2002), while the other raises doubtsabout their effectiveness in highly dynamic envi-ronments (Eisenhardt and Martin, 2000; Schreyöggand Kliesch-Eberl, 2007). Integrating these views,I found evidence for an inverse U-shaped con-tingent relationship where the effect of dynamiccapabilities on competitive advantage is highest inmoderately dynamic environments but lower underlow and high levels of environmental dynamism.


198 O. Schilke

Interestingly, the multi-group analyses (seeAppendix S1) showed that the positive effects ofthe two capabilities on competitive advantage werestill statistically significant in the high dynamismsubgroup. This finding appears to contradict Eisen-hardt and Martin’s (2000) and Schreyögg andKliesch-Eberl’s (2007) argument that dynamiccapabilities do not confer a competitive advantagein these settings. It suggests a less extreme positionin that these capabilities can be strategically valu-able even in high velocity environments, possi-bly because an inventory of established changerepertories can indirectly facilitate novel actionby providing the fodder for new recombinations(Levinthal and Rerup, 2006; Wirtz, Mathieu, andSchilke, 2007). Nonetheless, it is important to notethat the efficacy of dynamic capabilities decreasedsignificantly when moving from medium to highenvironmental dynamism, consistent with this arti-cle’s argument that they exert their relativelystrongest positive impact when dynamism is atintermediate levels.

This study contributes to research on dynamiccapabilities in several ways. First, it providesempirical support for the notion that dynamiccapabilities, like most ways of organizing, shouldnot be regarded as a universal, one-fits-all solu-tion. The study’s findings help delineate bound-ary conditions for dynamic capabilities theory—animportant precondition for any theory to moveforward. Second, the study establishes that envi-ronmental dynamism plays a key role in thelink between dynamic capabilities and competi-tive advantage. The article therefore contributesto answering ‘“under what conditions does thepresence of DC in firms generate competitiveadvantage?”: arguably one of the most interest-ing questions in the field of strategic managementtoday’ (Verona and Zollo, 2011: 537). Rather thanfocusing on dynamic contexts only, the study’smulti-industry design allowed for contrasting theefficacy of dynamic capabilities in settings withvarying dynamism. This research, thus, heeds callsfor empirical studies that ‘explicitly compare theeffects of similar dynamic capabilities in two ormore clearly distinct environmental conditions’(Barreto, 2010: 276). Results indicate significantdifferences among these settings, underlining theimportance of considering the degree of envi-ronmental dynamism when making claims aboutperformance implications of dynamic capabilities.Overall, this study thus helps reduce ambiguities

regarding the role of environmental dynamism inthe dynamic capabilities framework (Zahra et al.,2006). Third, this work makes a theoretical con-tribution by integrating existing theorizing on thecontribution of dynamic capabilities under varyinglevels of dynamism. I acknowledge both the costargument (Winter, 2003; Zahra et al., 2006; Zolloand Winter, 2002), which suggests that stable envi-ronments may not provide sufficient opportunitiesto cover the costs of d

Documents

ON THE CONTINGENT VALUE OF DYNAMIC CAPABILITIES FOR … · 2014. 1. 14. · on the relationship between dynamic capabilities, environmental dynamism, and competitive advan-tage. Integrating