Embed Size (px)
Citation preview
Strategic Management JoumalStrtii. Ms'iii. ./.. 26: 259-^276 (2(K)5t
Published online in Wiley InlerScience {www.imerscience.wiley.coni). DOI: U). 1 (H)2/smj.45O
MARKETING AND TECHNOLOGY RESOURCECOMPLEMENTARITY: AN ANALYSIS OF THEIRINTERACTION EFFECT IN TWO ENVIRONMENTALCONTEXTSMICHAEL S O N G ; CORNELIA DROGE,^* SANGPHET HANVANICH^ andROGER CALANTONE^' The Bloch School, University of Missouri, Kansas City, Missouri, U.S.A.; TEMA,Eindhoven University of Technology, The Netherlands' Eli Broad Graduate School of Management. Michigan State University, East Lansing.Michigan, U.S.A.^ Williams College of Business, Xavier University, Cincinnati, Ohio, U.S.A.
The dynamic capabilities perspective posits that a Jinn ccm Icverafii' the performance impactof fxistitiji resources through resource conjifiunilion. complftnenfarity. and inte^rotion. hut littleenipiriciil rc.veiinh addresses these issue.\. We inve.'iti^ute the effects itn perfornumce of nmrketinficapabilities, technolojiical capubiUties. atui their coniplenicntarity I interaction), and whetherthese effects are moderated by low vs. hi}>h technological lurhulencc. Results from .SfCM two-f>roup analyses (with ctmtn>ls) show thai both main effects positively impact perfonnance inboth environmental contexts. However, f / i their ititeraction effect I'.v .\ii;nijicant only in the high-turbulence environment: (2) the marketing-related main effect is lower in the high-turbulenceenvironment: and (3) the main effects of technolof-y-related capabilities are the .fame in bothenvironments. Our research sugffe.sts that the syncrgistic perfonnatice itnpact of complementarycapahilities can he substantive in particular environmental contexts: while svneti^istic rentscannot alwavs he obtained, it ispos.sihie to leverage e.xi.stin}' resources tluounh complementarity.Copyright © 2005 John Wiley & Sons. Lid.
INTRODUCTION
The relationships between resources (or capabil-ities) and firm performance have attracted muchresearch interest, but we still know relatively lit-tle about why some firms successfully use theircapabilities while others do not (Helfat. 20(X)). Theextant literature suggests that superior performance
Keywords: dynamic marketing/technological capabilities;resource-ba.sed theory: inieraciion cffccl; SEM' CoiTcspondence to: Cornelia Drogc. Eli Broad Graduate Schtwlof Management. Michigan Slate tJniversiiy. N.17() Nonh Busi-ness Complex. East Uinsing. MI 48824-1122. U.S.A.E-mail;
can come from resource uniqueness (e.g.. Bar-ney. I99I). from reconfiguration and integrationof existing resources (e.g.. Risenhardt and Mar-tin. 2000; Teece. Pisano. and Shuen. 1997). and/orfrom ihc ability to respond appropriately to thesurrounding environment (e.g.. Mintzberg, 1987;Pfeffer and Salancik. 197S; Tan and Litschert.1994). Our study aims to contribute to this lit-erature by focusing on two issues that are rel-atively neglected: (1) the performance impact ofthe ititeraction of capitbilities (in addition to maineffects); and (2) the ilijfcrential impact of capa-bilities and their interaction in different environ-ments. The former addresses whether complemen-tary capabilities have synergistic effects, while
Copyright © 2005 John Wiley & Sons. Ltd. Received 17 Octoher 2002Final revision received 17 September 2004
260 M. Song et al.
the tatter specifies environmental conditions underwhich both main and synergistic effects can beexpected. Specifically, we investigate the relation-ships to performance of marketing-related capa-bilities, techno logy-related capabilities, and theirinteraction in two environmental contexts; high vs.low technological turbulence. We tap perfonnanceby considering profit, sales, and ROI relative toobjectives.
Technological-related capabilities have beenshown to enable firms to achieve superior per-formance (e.g.. Clark and Fujimoto. 1991; Pisano.1994). Likewise, marketing-related capabilitieshave been established as important resources formarket-driven organizations (Day, 1990. 1994).The focus of this paper goes beyond the impactof these main effects; rather, we scrutinize the rel-atively unknown and under-researched impact oftheir joint presence (their interaction) under differ-ent environmental conditions. Thus our first broadgoal is to contribute to the literature by enablingan evaluation of complementarity in capabilitiesthrough our modeling of interaction. However, theanalysis of construct interaction effects is still inits infancy (Jaccard and Wan. 1996). and thus toaccomplish this goal we use a little-used method-ology to model interaction constructs in structuralequation mtxleling (SEM). We hope that thesemethodological a.spects wilt encourage more inter-est in construct interaction effect analysis.
Our second broad goal is to examine the mod-eration of teehnological turbulence (a fomi ofenvironmental uncertainty) on the relationships toperfonnance (.)f both main and interaction effects.Various degrees of technological turbulence, withassociated rates of product or process obsolescenceand new product introduction, characterize the cur-rent competitive environment of high-tech indus-tries. Surprisingly, little research empirically testswhether, for example, the performance impact oftechno logy-related capabilities is greater in highas compared to low technologically turbulent envi-ronments. We address these issues in the followingresearch question: Is perfonnance affected difjer-entially hy each individual capability (the m;u-ket-ing or technology capabilities main effects) and/orby their joint presence (the interaction of thesecapabilities), depending on the level of this tech-nological turbulence?
We begin our paper with the development ofmain effects, interaction effect, and nnxlerationhypotheses, and then test them using new product
commercialization joint ventures (JVs) as a setting.New product commercialization is not only cru-cial for the materialization of technotogy-retatcdcapabitities (Page, 1993). but is atso ttie stage inthe new product development process where theinteraction between technology-related capabilitiesand marketing-related capabilities is most likelyto occur. We used Joint ventures because they are•firms" born of strategic alliances whose very pur-pose may be providing firms with access to com-plementary a.ssets (Harrison et al., 2001; Kogut,1988). This allows us to focus on relatively nar-row firm capabilities in a context hospitable for theempirical testing of our hypotheses.
RESOURCES, CAPABILITIES, ANDPERFORMANCE
In the following sections, we develop six hypothe-ses that, as a .set. specify different relatitmshipsto performance of marketing-related capabilities,tech no logy-related capabilities, and their interac-tion. Differences arc hypothesized to be engen-dered by technological turbulence. Our model alsospecifies three control variables: market growth,relative costs, and industry. Grounded in the re-source-based view, the model's hypotheses aresummarized in Figure 1.
Resource-based theory: A brief summary
Re sou ice-based theory views a (inn as a uniquebundle of tangible and intangible resources andemphasizes the protection of firm core competen-cies comprising these resources. Several authors(Bamey, 1991; Day and Wensley, 1988; Praha-lad and Hamel. 1990; Wemerfelt. 1984) haveexpanded the seminal work of Penrose (1959).Resources include all a.ssets. capabilities, organiza-tional processes, firm attributes, information, andso on controlled by a firm and enabling the firm toconceive of and implement strategics that improveefficiency and effectiveness (Barney, 1991), Finncompetitive advantage is rooted in resources thatare valuable and inimitable, and the firm's survivallargely depends on how it creates new re,sources,develops existing ones, and protects its core com-petencies (Day and Wensley. 1988).
The resource-base view of the firm is notrestricted solely to examining internal resources,however. Several authors recognize that many
Copyright C 200S John Witey & Sons. Ltd. Strat. Mgmt. J., 26: 259-276 (2005)
Marketing and Technology Resource Complementarity 261
Technology-Related
CapabilitiesHI : In both environments, ihe greater the technolog\-related capabilities, the better the J Vs performance.
H4 (H4alt): The strength of the relationship between technology-related capabiliiies and .IV performance is greater (or lower forH4alt) in an environment characterized by high technologicalturbulence than in an environment eharacterized by lowtechnological turbulence.
Marketing -Related
Capabitiiics
Performance(Relative loObjectives)
H2: In both environments, the greater the marketingrelated capabilities, the better the JV's pertbrmance.
H5: The strength oCthe relationship betweenmarketing-related capabilities and JV perlbnnanceis lower in an environment characterized by hightechnological turbulence than in an environmentcharaeterized by low technological turbulence.
HJ: ln both environnient.s, marketing-relatedcapabilities and technology-relatedcapabilities will interact to positively affectJV performance (in addition to the mainefFects of each capability on performance).
Marketing XTechnologyInteraction
Effcet
116: The relationship to .IV performance of the interactionofniarketing and technology related-capabilities is greater ina high technologically-turbulent environment than in a lowtechnologically-turbulent environment.
Figure I. Theot^tical model of marketing and technology resource complementarity in the two environmental contexts
essential resources and capabilities lie outside thefirm's boutidatnes (Doz and Hamel, 1998). Grant{1991). for example, stated that when internalresources are unavailable, outsourcing should heconsidered, and Das and Tetig claim that hy joiningforces with other firms a firm can gain 'otherwiseunavailable competitive advantages and values'(Das and Teng. 2000: 36). Ititegration of tangi-ble or intangible resources from participating firmsprovides a joint venture or alliance with strategic
rents that are achieved not necessarily because ithas better or more resources, but rather because theventure's distinctive competence involves makingbetter use of joint resources (Penrose, 1959).
Marketing vs. technology-related capahilities:Two key resources
There are many ways to define 'capabilities.'Collectively, capabilities are defined as complex
Copyright © 2005 Jiihn Wiley & Sons. Lki. Strat. Mgmi. J.. 26: 259-276 (2005)
262 M. Song et al.
bundles of skills and accumulated knowledge,exercised through organizational processes, thatenable Hrms or joint ventures to coordinate activ-ities and make use of the asset (Day. 1994).In this research, we focus on marketing-relatedcapabilities vs. technology-related capabilities injoint ventures (JVs). Although established throughcooperation between firms, a JV is considereda separate legal entity or a 'firm' in its ownright (Murray and Siehl. 1989: Park and Ung-son, 1997). Therefore, technology and marketing-related capabilities are regarded as 'firm'-leveltraits. Marketing-related capabilities are tho.se thatprovide links with customers; they enable JVs tocompete by predicting changes in customer pref-erences as well as creating and managing durablerelationships with customers and channel members(Day, 1994). Technology-related capabilities arethose that develop and produce technology; the,seenable response to the rapidly changing techno-logical environment (Wind and Mahajan. 1997).Thus both capabilities are idiosyncratic resourcesthat can provide competitive advantage (Barney,1991; Peteraf. 1993; Wemerfelt, 1984). There-fore, according to the resource-based perspective.Hypotheses 1 and 2 are hypothesized. Neither isnew, but both are necessary for model complete-ness.
Hypothesis !: The greater the technology-relat-ed capahilities. the better the JVs performance.
Hypothesis 2: The greater the marketing-relatedcapahilities. the hetler the JVs peiformance.
Joint ventures are not only an effective meansto share complex capabilities among the venturepartners (Kogut, 1988; Mowery, Oxley. and Sil-verman, t996). but atso an attractive vehicte forenhancing hrm capabitities (Madhok. 1997). Capa-bilities can be divided into complementary andsupplementary capabilities; complementary capa-bilities are those that combine effectively withthose the firm already has. whereas supplementaryresources are those that serve the same functionsas the ones the firm already has (Wernerfclt. 1984).Integrating marketing capabilities and technologi-cal capabilities should lead to better performancebecause it is a complementary rather than sup-plementary combination. Such integration recon-figures competencies, reduces the resource defi-ciency, and generates new applications from those
resources (Kogut and Zander, 1992; Teece et al.,1997; Woodcock. Beamish, and Makino. 1994).Complementary resource combinations will alsocontribute to the J V s balance of power: balanceis crucial for JV success (Bucklin and Sengupta,1993; Heide. 1994) and stems from the equalresource dependence of both parties (Emerson,1962; Gaski, 1984). Therefore it is hypothesizedthat:
Hyptithesis 3: Marketing-related capahilitiesatid techtiology-related capahilities will inter-act to positively affect the JVs performance (inaddition to the main effects of each capahilityon perfonnance).
The moderating effect of low vs. hightechnological turbulence in the environment
Consideration of the environment is important tothe analysis of firm resources and performancesince different environments imply different val-uations of resources (Fenrose, 1959). In particular.JVs are often chosen in order to respond to thecontinuing global technologically titrbtdcnt envi-ronment (Achrol, 1991; Collis. 1991). Such JVsusually seek to enhance strategic advantage byleveraging critical capabilities (sueh as technology-related and marketing-related capabilities) and byimproving flexibility in response to technologicalchange (Achrol, 1991). According to the dynamiccapabilities model, and more broadly the resource-based view, uncertain and turbulent environmentshelp firms achieve competitive advantages hecauseuncertain turbulent environments increase causalambiguity and. as a consequence, the ability toimitate resources or combinations of resourcesdecreases (e.g.. Eisenhardt and Martin, 2(X)0; Lipp-man and Rumelt. 1982: Noda and Collis, 2()()l). Inhighly turbulent environments, the JV can deployresources from each participant in order to respondto changing ct)nditions: thus, the way the JV usesresources and the joint capabilities to be developedwill not be static. This is difficult for competi-tors to imitate in a timely fashion. On the otherhand, when the environment is relatively unchang-ing and predictable, competitors can see clearlywhich resources and combinations of resources arevaluable to the business, and these can be imitatedbecause time is not of the essence.
Copyri^t ® 2005 John Wiley & Sons, Ltd, Slrat. Mumt. J.. 26: 259-276 (2005)
Marketing and Technology Resource Complementarity 263
Consider first technology-related capabilities. Ahighly technologically turbulent environment ischaracterized by a short cycle of technologicalinnovation and obsolescence. In high turbulence,technology-related capabilities (such as innova-tion) should enable a JV to shape or react tothese environmental conditions (Kotabe and Swan,1995). For example, the timety intrtxiuction of newproducts to replace obsotete products may becomecrucial to firni success (Wind and Mahajan. t997).Therefore, the relationship between technology-related capabilities and perl'ormance in a hightechnologically turbulent environment should begreater than this relationship in a low-turbulenceenvironment (i.e.. the betas will not be the same).
It can. however, also be counter-argued thatembedded technological capabilities may lead toincumbent inertia when the environment becomestechnologically turbulent (Lieberman and Mont-gomery, 1988). Deeply embedded knowledge andskill sets can actually create problems if firmsfail to fill the gap between current technologicalenvironmental requirements and their core tech-nological capabilities, thus creating core rigidities(Leonard-Barton, 1992). Technological changescan therefore either enhance or destroy the exist-ing firms' technological competencies (Tushmanand Anderson. 1986). We address this paradox byproposing both Hypothesis 4 and 4alt:
Hypothesis 4: The strength of the relatiotiship(i.e.. the heta) hetween technology-related capa-hilities and performance is greater (Hypothe-sis 4alt: lower) in an envirotiment character-ized hy high technological turbulence than inan environment characterized hy low technolog-ical turhulence.
Next, consider marketing-related capabilities,which enable JVs to gain and use marketintelligence about exogenous market factors thatinlluence cunent and future customer needs. Inthe high technologically turbulent environment, therole of tnarketing-reldled capabilities in generatingperformance may be downplayed, particularlyin the situation where the whole industry isaffected by rapid technological change. In sucha situation, the importance of close relationshipswith customers or among supply chain membersmay decrease, whereas the importance of newproduct intrtKluction increases. Customers may notbe able to help firms innovate (although they can
be used to test products), and thus technology-related capabilities must assume a dominant rolein performance responsibilities. Therefore, wehypothesize:
Hypothesis 5: The strength of the relationship(i.e., the heta) hetween marketing-related cafya-hilities and performance is lower in an environ-metit characterized hy high technological turhu-lence than in an environment characterized hylow technological turhuletice.
In a high technologically turbulent environment,JV partners will not be able to predict futurechanges. In such a situation, diversity in capa-bilities should provide JVs with more diversifiedideas, which should lead to better risk manage-ment and higher success. As such, the effect onperformance of the complementarity of marketing-related and technology-related capabilities shouldbe greater in a high (vs. low) technologically tur-bulent environment. Therefore, we propose:
Hypothesis 6: The relationship to ,/V perfor-trumce of the interaction of marketing atul tech-nology-related capabilities is greater in a hightechnologically iiirhulftit euvirot\ment than in alow technologically turhulent environment.
METHOD: SAMPLE ANDMEASUREMENT
Sample and procedure
We tested our hypotheses using survey data. Theinitial sampling frame was obtained from a com-mercial listing of U.S. joint ventures formed be-tween 1990 and 1997. After eliminating firms forwhich the questionnaire was inappropriate, theoverall frame had 971 JVs. In administering themail survey, we followed the modified total sur-vey design method (Diltman, t978), and obtained466 usabte responses (response rate = 48%). Acomparison of the responses from two maitingsrevealed no systematic differences in the studyvariables.
The respondents consisted of 79 presidents; 214vice-presidents of marketing or directors for mar-keting operations; 187 vice-presidents of R&D ormanufacturing; and 61 others. Informant tenure
Copyright © 2005 John Wiley & Sons, Ltd. Strut. Mgmt. J.. 26: 259-276 (2005)
264 M. Song et al.
levels with the JV averaged 6 years. The aver-age number of employees in the JVs was 792,with a range of 57-1650 (this is an indica-tor of JV size). The industries represented were:Chemicals and Related Products; Electronic andElectrical Equipment; Pharmaceutical, Drugs, andMedicines; Industrial Machinery and Equipment;Telecommunications Equipment; Semiconductorsand Computer Related Products; Instruments andRelated Products.
Measurement of key model constructs
Before collecting data, we conducted four in-depthcase studies to validate measures. Table I presentsthe wording and scale points of key model Viiri-ables. Cumulative normal probability plots demon-strated that each of these measures was normallydistributed. Appendix I contains the complete cor-relation matrix.
Respondents were required to rate the market-ing-related capabilities and technology-related
capabilities of the JV. The marketing-related capa-bilities, focusing on market sensing and exter-nal linking capabilities, were developed from Day(1994), The techtiology-related capabilities, focus-ing on technology development, new productdevelopment, and manufacturing processes, werealso drawn from Day (1994). In addition to thesetwo latent independent constructs, we also havethe following independent variables as controls:(1) market growth, the average annual growth ratein percentage of total sales in the JV's princi-pal served market segment over the past 3 years;(2) relative costs, the JV's average total operat-ing costs in relation to those of its largest com-petitor in its principal served market segment;and (3) ituiustry (six dummy variables represent-ing seven industry groups).
Finally, the dependent construct perfonnancerelative to profit, sales, and ROI objectives wasmeasured on I l-point scales anchored 'lowV'high.'Using perceived performance scales relative toobjectives permits comparisons across firms and
Table 1, Measurement items and response formats
Constmcl and response formal Meiisuremeni ilerns
Marketing-related capabilities (MKT)Please evaluate how well or poorly you believe ihis
joint venture performs ilie specific activities orpossesses tfie specific capabilities relative to yourmajor competitors. (1 l-point scale with anchors;0 = Much worse than your major competitors:10 = Much better than your major competitors)(adapted from Day. 1994)
Tecbnoloi^y-related capabilities (TECH)Please evaluate how well or poorly you believe ihis
joint venture performs the specific activities orpossesses the specific capabilities relative to yourmajor competitors. (1 I-point scale with anchors:0 = Much worse than your major competitors;10 = Much beuer than your major competitors)(from Day. 1994)
Technologically-turbulent environmentPlease indicate the degree to which you agree or
disagree with the following statement regardingthis joint venture (1 l-point scale with anchors;0 = strongly disagree: 10 = strongly agree)
Overall perfonnancePlease rate the extent to which this joint venture (JV)
has achieved the following outcomes. (1 l-pointscale with anchors: 0 = low: Ht — high)
Customer-linking capabilities (i.e , creating and managingdurable customer relationships)
Market-sensing capabilities (predicting changes in customerpreferences)
Channel-bonding capabilities (creating durable relation.shipwith channel members such as wholesalers, retailers)
Technology development capabilitiesManufacturing processesNew product development capabilities
The technology in our industry is changing rapidlyTechnological changes provide big opportunities in our
industryIt is very difficult to forecast where Ihe technology in our
industry will lie in Ihe next 2-3 yearsTechnological developments in our industry are rather
minor (R)
Overall profit margin relative to the JV's objectiveOverall sales relative to the JV's objectiveOverall RO! relative to the JV s objective
Copyright © 2005 John Wiley & Sons, Ltd. Slrat. Mgmt. J.. 26: 259-276 (2005)
Marketing and Technology Resource Complementarity 265
contexts (such as across particular industries, cul-tures, time horizons, economic conditions, andexpectations of parent firms). The managers in thecase studies preferred subjective to objective mea-sures because the latter are often confidential. Theliterature shows that subjective scales are widelyused and that there are high correlations betweensubjective and objective firm performance mea-sures. Finally, note that pertbrmance objectivesare determined with capabilities in mind, and thusmeasuring actual performance relative to objec-tives creates a ptitential bias against finding sig-nificant effects.
Cla.ssification of high vs. low technologicalturbulence
Perceived technological turbulence refers to thestate of technology in the industry, the rate ofchange in technology, and the JV's inability toaccurately forecast the changes in the technology(Downey and Siocum. 1975: Milliken. 1987). JVswere classified in two steps. First, three researchersassessed the technological environments by label-ing as 'high' those with the following characteris-tics: strong network externalities (Xie and Sirbu.
1995): high uncertainty: rapid changes in industrytechnology standards; short technology life cycles(less than 2 years); and faster development cycletime (less than I year for typical new products).Majority rule resolved disputes. This classifica-tion scheme is consistent with Song and Montoya-Weiss (2001). Second, we calculated the samplemean for the composite score of the perceivedtechnological turbulence scales (Table I). Basedon this mean score. JVs were sorted into 'high' or"low." For a JV to be included in the fmal usablesample (n = 466). it had to have the same classi-fication from bt)th methods (19 JVs were droppeddue to mismatch). The result was 249 JVs in thehigh and 217 JVs in the low technological turbu-lence group.
Equivalence of measurement across high vs.low groups
The equivalence of measurement across groupswas assessed by the set of hierarchical tests as out-lined by Bollen (1989) and summarized in Table 2.The initial model (Model I), without constraintsacross groups, provided a baseline chi-square. Theresults showed a good model fit (X'dhi = 35.99;
Table 2. Analysis of the measurement model across environmental groups
(A) Twn-gmup analyses: icsls for equivalence nf measiircmcnl ami discriininani vnlidiiy
Measurement mcKfel Goodness of fit Test of hypolheses
Model Ml: Baseline modelModel M2: Factor loadings modeled
invariant
Model M3: Factor loadings ami errorvariance modelled invariant
Mtidcl M4: Factor loadings invaiiant {iritlcorrclalion belween marketing-related andtechnology-related capahiliiies sel to 1
., = 35.99./; = 0.00o, = 43.64. p = 0.00
= 59.36, p = 0.00
= 58.31. p =0.00
Test for loading invarianceModel 2-Model I:^X\,. = 7.65, n.s. at 0.05Test lor invarianceModel 3-Model 2:A/-,., = 15.72. sig. at p < 0.05Test for discriminanl validityModel 4-Modc! 2:Ax",!, = 14.67, sig. at p < 0.05
(B) Moasurcmcnt model wilh I'actor loadings constrained ecjual across groups
Measurement model(constraints equal)
Unstandardi/ed solution(/-value in parentheses)
Coninu)n metric completelystandardized solution
(MKT) ;(MKT) X(MKT)(TECH)
(TECHI
1.001.03" (11.65)0.48" (9.94)
1.001.02" (16.39)1.02- (16.36)
0.840.800.520.820.800.78
" Significant at /J < O.OI
Copyright S 2005 John Wiley & Sons. tJt Sirai. Mgrni. J.. 26: 259-276 (2005)
266 M. Song et al.
RMSEA = 0.07). The second step (Table 2A) wasto constrain the factor loadings equal: the non-significant difference in chi-square between thismodel (Model 2) and the baseline model (ModelI) indicated that ihe factor loadings were invariant(Ax',A> = 7-65. n.s. al p < 0.05). Third, we testedIhe equality of the error variances of the latent vari-ables (Bagoz/i and Edwards. 1998). A significantdecrease in chi-square between Model 2 and Model3 (Ax'iM = '5-72. p < 0.05) indicated differenterror variances. Thus the measurement model wask loading invariant only. This X invariant model(Model 2) was used in subsequent analyses.
An examination of the loadings of Model 2(Table 2B) indicated that a substantial amount ofvariance was captured by the latent constructs: allloadings were highly significant and only one slan-dardtzed loading was below 0.7, showing strongconvergent validity. The tesi of discriminant valid-ity (Table 2A. MIKJC! 4) involved comparing chi-square values of models that either free or con-strain the coiTciation between constructs to 1. Thedecrease in chi-square was significant (A / ' , , , =14.67. significant at p < 0.05). supporting dis-critninant validity.
METHOD: INTERACTION EFFECTESTIMATION IN SEM
Our approach to interaction effeet analysis usingSEM. outlined below and detailed in Appendix 2 isin line with that first suggested by Kenny and Judd(1984). It involves first centering the raw scores.The measurement equations of FM (marketing-refaled capabilities) and F^ (technology-relatedcapabilities) are, in deviate form:
and
M, =
T, =ki,F,
-\- (1)
(2)
interaction. Therefore constraint #2. detining thepath coefficients (A) between interaction construct(FMFJ) and its tnultiplicative indicators (MT).. is:
with errors of the product indicators as:
fMiTj = (A.M,F«er/) + {^rj'v^M/) + ^M.^r/ (5)
The residual variances of interaction indicators are:
(6)
and constraint #.^ defming the residual variancesof interaction indicators, is:
(7)-t-
From Equations 6 and 7 it can be shown (hat allpaths between error terms of multiplicative indi-cators must be freed except where there is novariance sharing. This set of fixed paths establishescon.straini fM (described in detail in Appendix2). The final step is to establish the covariancesbetween the interaction and the other latent con-structs: these are zero for nonnally distribttted andtnean centered variables (see Appendix 2). There-fore the final con.itraint #5 is:
— <i>FT.FMFT — 0 (8)
Then the variance of an interaction latent constructis as follows (con.straint 01):
The five constraints outlined above show that agiven multiplicative indicator is a function of themeasurement error of the component parts of theinteraction tertn. An analysis strategy not tak-ing into account this complex function will causepoor model fit and erroneous results. SEM resultsfrom analysis of ititeraction withoui these con-straints (available from the authors) substanllallydepart from results wiili these constraints (reportedbelow). Furthermore, the results depart from thoseof regression procedures such as OLS (demon-strated below ;ind in Table 4).
= Var(FM)Var(Fr) -I- Cov(F«. (3)
The second step is to establish the path coefficients(i.e., A) and the error variances (i.e., eMtrj) for the
RESULTS
The results reported are from the I.ISREL modelshown in Figure 2. We used invariant factor load-ings, justified by the measurement tests. Also,
Copyright © 2005 John Witcy & Sons. t,ld. Siral. Mgmi. J.. 26: 259-276 (2005)
Marketing and Technology Re.wurce Complementarity 267
Control Variables:Growth, Cost, and Industry (six dummy variables to represent seven industries)
firowih1 i i i
Flee
IVIiirketing-Rclatcil
Ca pa bi lilies
" We let technological capabilities and marketing capabilities covarj'.
" Some paths between the errors of the product terms (constraint M) are not shown in ihis figure.
Figure 2. LISREL model of marketing and technology resource complementarJiy (with control variables) in twoenvironmenlal contexts
based on constraint #2, the factor loadings of theinteraction constmcl are functions of the factorloadings of the main latent constructs.
SEM analyses and hypothesis testing
We first tested the equality of the control vari-ables" effects across groups. When these paths
were constrained equal, chi-square did not changesignificantly from ihe baseline Model I (Table 3).Thus the effects of control variables were not sta-tistically different across two groups.
The SEM results in Table 4 from the base-line model showed that the paths to perfor-mance from marketing-related capabilities and
Copyright © 2005 John Wiley & Sons. Ltd. Strat. Mgmt. J.. 26: 259-276 (2005)
268 M. Song et al.
technology-related capabilities were highly sig-nificant in both low and high technologi-cally turbulent environments (low: XIK-PKRK =0.53. / = 6.58: KMKT^PKRF = 0.61. I = 8.39; high:/rEc^PERF = 0.59, r = 9.64; KMKT^PERF = 0.29,/ = 4.58). However, the path from the interac-tion effect to performance was signilicant onlyin the high technologically turbulent environment
(low: j'lNx-piiHH = —0.03. I = —0.84. n.s.; high:ViNX-PBRF = O.IO. r = 4 . 1 6 ) . These results thusprovide support for Hypotheses 1 and 2 and partialsupport for Hypothesis 3.
Next, we tested the hypotheses that the pathcoefficients to pertbrmance from market ing-relatedcapabilities, technology-related capabilities as wellas the interaction effect are different across the
Table 3. Two-group anjlysis: hypotheses testing
Structural model Goodness of fit Test of hypotheses
Model 1: Baseline model (factor loadingsinvariant)
Model 2: Factor loadings f/n(7 path x',m, = 1575.21. p =0 .00coefficients between control variables andpedbrmance invariant
;- ; ,, = 1567.19. p = 0.00 Test for Hypotheses 1. 2. 3
Model 3: Factor loadings and pathcoefficients KMKI-I>KR. Kirr-H^H. and/(Nx-PER invariant
Model 4: Factor loadings and pathcoefficient ) n< ntR invariant
Model 5: Factor loadings and pathcoefficient KMKr..n:B invariant
MtMJei 6: Factor loadings and pathcoefficient } iNx-n:R invariant
'^\^^^^ = 1587.03. p = 0.00
;2 , ,, = 1567.46. /) = 0.00
:\m. = 1578.79. p = 0.00
r.m^ = 1576.23, /; = O.tX)
Test for equalities across groups ofthe control variables onperformance
Model 2-Model I:A/-,g, = 8.02, n.s. at 0.05
Model 3-Modei 1:A/-,,, = 19.84, sig. at p < 0.05
Test for Hypotheses 4/4aItModel 4-Model I:A/-,,, =0 .27 . n.s. at 0.05
Test for Hypotheses 5Model 5-Mode! I:Ax',n = ' 160 . sig. at p < 0.05
Test for Hypotheses 6Model 6-Modei I;Ax'ii, =9 .04. sig. at p < 0.05
Table 4. Results of OLS vs. structural equation model analysis with control variables
Path coeflicients(r-valuc in parentheses)
J^li-PEKF
J MKl —PtRF
MNX-PEBF
Control variablesKOHOWTH—PERF
yCOST-PURF
VEr.KC-PERF
ypllAR-reRF
yiNDM-PI-Ri:
/FTLR^HiRF
ySfiMI-PfiRF
" Significant al /J < 0.05
Copyrighl © 2005 John Wiley
OLS
Low tech.turbulence
0.52'(2.69)O.57*(3.2O)
-0.02 (-0.36)
0.00 (1.00)0.09 (0.97)
-0.50 (-0.73)-0.76 (-1.19)-t) .23(-0.36)
0.63 (0.97)t).23 (0.35)
-0.06 (-0.10)
& Sons. Ltd.
High tech.turbulence
0.04 (0.28)-n.13 (-0.85)
O.O8-(3.27)
0.01-(2.52)0.01 (0.13)0.05 (0.07)0.74(1.19)0.90(1.34)0.73 (1.22)0,20 (0.33)0.31 (0.60)
SEM
Low tech.turbulence
n.53*(fi.58)0.61-(8.39)
-0.03 (-0.84)
0.00 (0.92)0.05 (0.49)
-0.48 (-1.14)-0.54 (-1.32)-0.01 (-0.01)
0.67(1.56)0.41 (0.93)0.06(0.17)
High tech.turbulence
().59-(9.64)().29-(4.38)0.10-(4.16)
0.01'(2.68)0.02 (0.20)0.06 (0.13)0.89-(2.12)1.06'(2.36)0.84-(2.23)0.28 (0.70)0.35(1.07)
Srrat. A
A Result
Test of high vs. low
InvariantSignilicanily differentSignificantly different
InvariantInvariantInvariantInvariantInvariantInvariantInvariantInvariant
•Iftmt. J.. 26: 239-276 (2005)
Marketing and Technology Resource Complementarity 269
two environments (Model 3 in Table 3). Theresult (Ax",3, = 19.84, /) < 0.05) demonstrateddilTerences. and additional tests identified whichpiuhs dilYcred. The first test was to assess theinvariance of >^c-PHRf: by constraining the pathlo be equal across groups (Model 4, Table 3).Km--rF.Rh tested equal across groups (Ax'm =0.27. n.s. al 0.05) and ihis supptmed neitherHypothesis 4 nor Hypothesis 4ah (Model 4 inTable 3). The second test was tor invariance ofthe path from marketing-related capabilities loperformance. The significant difference in chi-sqiuuc supported Hypothesis 5 in that KNIKT-PF-RI'
in the high technologically turbulent environmentwas significantly lower than KMKT -P[-.R(- in the lowtechnologically turbulent environment (A/ ' , , , =11.60. p < 0.05). Finally, the test of invarianceof the path coefficient from the interaction to per-formance (Model 6 in Table 3) showed a sig-nificant difference in chi-square (A/- , , , = 9.04.p < 0.05), supporting Hypothesis 6 that the inter-action effect in high turbulence was greater thanthe one under low turbulence.
In addition. Table 4 compares our SEM resultswith the results from ordinary least square (OLS)regression. Our purpose is to demtmstrate that OLSresults can lead to substantively different conclu-sions. For OLS analysis, we took the niean of theindicators of each construct (thus there is no "mea-surement model' as in SEM: e.g.. measurementerror is not explicitly modeled) and the ordinarymultiplicative interaction. The results from OLSdiffer from SEM analysis. For example, in thehigh technologically turbulent environment, nei-ther main effect is significant in OLS. OLS esti-mates, being conditional on other mcxiel variables,changed in the high tech turbulence group becauseof" tlie significance of the interaction in this group.All beta estimates will differ across SEM and OLSbecause OLS does not account for psychometricproperties of the measurement model (both con-structs and interaction).
The strength of the interaction effect and itsinterpretation
The strength of the interaction effect is reflected inthe difference between the squared multiple cor-relation (similar to R- in OLS) of models with-out vs. with interaction (Jaccard and Wan, 1996).The latter was modeled by fixing the value ofthe path coefficient between the interaction effect
latent variable and performance to zero. However.given that only the interaction effect in the hightechnologically turbulent environment was signifi-cant, the effect in the low technologically turbulentenvironment was not examined. In the high tech-ruilogically turbulent environment, the square mul-tiple correlations without and with the interactionwere 0.42 and 0.51 respectively. This means thatmarketing-related and technology-related capabil-ities together accounted for 42 percent of vari-ance in perfonnance. while the interaction effectaccounts for 9 percent. However, this is a some-what crude index.
When an interaction effect is statistically signif-icant, it should be further analyzed and interpretedas a conditional effect on the main effects (Jaccard.Turrisi, and Wan. 1990). Specitically, the effectof marketing-related capabilities on performance,at a given level of technology-related capabil-ities is: /?MKT al Vtei.' = yMKT-I'tRl- "f KlNX-PEKr-V'ttclwhere V,,. is a specific value of technology-relatedcapabilities and y are path coefficients as dis-cussed above (similarly: hjvx: .i vmki = yvw^paw +KiNx-PKRi- mki)- When assuming mean deviateform (as in this study), the mean of V is of coursezero. For instance, in the high technologicallyturbulent environment, an increase of marketing-related capabilities by one unit was estimated toincrease performance by 0.29 units, given thatthe JV has an aveiage level of tech no logy-relatedcapabiliiies. That is: /JMKT «. vi«: = yMKi-ciatF +I IN.X -I'l-RhKa ~ >'MKT--PBRF + ViNX—PERF(O) = 0-29
+ 0.10(0 = 0.29.When the values of the exogenous constructs
are not at their means, V can be obtained (ina standard deviation form) from the square rootof the variances. The variances of latent techno-logical capability and marketing capability con-structs are. respectively, 4.54 (/ = 10.06) and6.14 ( / = 8.50) in the low technologically tur-bulent environment and 7.13 (/ = I1.6I) and6.45 (/ = 9.52) in the high technologically tur-bulent environment. For example, when the levelof technology-related capabilities is "high" (suchas one estimated deviation above its samplemean), the effeet of market ing-related capabili-ties on performance (in the high technologicallyturbulent environment) can be calculated as fol-lows: /JMKT ai I'lec = XMKT-PKRF + KiNX-PERF-Kec =KMKT ^ PERF + YwK .nwV^ = 0.29 -I- 0.10(N/7 .13) = 0.56. For every unit that marketing-related capability increases, performance increases
Copyrighl © 21X15 John Wiley & Sons. Ltd. Sinii. Mgint. / . 26: 259-276 (2005)
270 M. Song et al.
by 0.56 units. This is an incremental increase of0.27 units when compared to the value when tech-nological capabilities are at the mean.
Using the same calculations, the effects oftechnology-reiated capabilities and marketing-related capabilities on pertbrmance in a lowicchnologically turbulent environment will alwaysbe 0.53 and 0.61 units, since the latent interactionconstruct is not statistically significant. In the hightechnologically turbulent environment, (1) theeffects of technology-related capabilities are 0.84.0.59, and 0.34 units, when the marketing-relatedcapabilities are high, at their means, and lowrespectively and (2) the effects of marketing-related capabilities are 0.56, 0.29, and 0.02 units,when the technology-reiated capabilities are high,at their means, and low respectively.
DISCUSSION
This research provided a contextually robust test ofdynamic capabilities and, more generally, resource-based theory, in the joint venture arena. We mod-eled the etfects on performance {profit, sales,and ROI relative to objectives) of (1) marketing-related capabilities. (2) technology-related capabil-ities, and (3) their interaction effect. The appro-priate constrained structural equation model wasused to test the hypotheses. Although our approachdoes not answer the question as to which specificlevels of investment in resources (i.e., capabilities)is best, it doe.s set the basis for synergy proposi-tion testing in a field that claims synergy throughcomplementarity but has not shown it empirically,in addition, the moderating effect of technologi-cal turbulence (low vs. high) was incorporated inthe theoretical model. Overall, our model providesIhe foundation for straightforward but powerfulmanagerial and theoretical guidelines without thepossibly misleading oversimplifications and with-out compromising the richness of the contextualsetting.
The main effects of marketing-related andtechnology-related capabilities
Results from two-group analysis showed that bothmarketing-related capabilities and technology-related capabilities were positively related toperi'ormance. These capabilities are the resourcesof the JV, atid, consistent with resource-based
theory, resources have positive performanceitnpact. From a managerial point of view, theresults confirm that JV performance can beenhanced by utilizing the right tnarketing andtechnology capabilities effectively.
The main effects of matketing-related and tech-nology-related capabilities on performance werepositive regardless of technological turbulence. Fortechnology-related capabilities, the strengths of therelationships to performance were equal (i.e.. thispath was not moderated by technological turbu-lence). We had expected a difference in the slopes,but this was not the case. For marketing-relatedcapabilities, the relationships were not the same inboth contexts: the strength of the relationship (i.e.,the slope) was greater in the low technologicallyturbulent environment (however, even in high tur-bulence, this main effect was positive: i.e.. it wasnot nil).
For managers, the implication is clear: care-ful management of capability deployment (i.e.,resource deployment) according to environmentalconditions is essential for maximutn pertbmiance.In our research, the performance impact of deploy-itig tnarketing-rel ated capabililies was greater in alow technologically turbulent environment, whilethe pertbrmance impact of deploying technology-related capabilities was the same across this par-ticular environmental characteristic. In low turbu-lence, the performance effects of marketing-relatedand of technology-related capabilities were vetysimilar: but with high turbulence, the effects ofmarketing-related capabilities (0.29) were not at allsimilar to the effects of technology-related capabil-ities (0.59). In general, managers and researchersfrequently fail to take into account the tiioderationeffects of environmental contexts, such as techno-logical turbulence as moderator.
The interaction of marketing-relatedcapabilities and technology-related capabilities
Re source-based theoi y claims that cotnpletnen-tary resources tnay enjoy synergistic performanceimpact, but this is rarely empirically tested. Thuswe modeled the interaction's effect on perfor-mance in addition to the main effects. We expecteda positive interaction effect in both groups and agreater beta in the high technologically turbulentenvironment, but the effect was significant only inthe high-turbulence environment. Clearly, resource
Copyright © 2005 John Wiley & Son:,. Ltd. Sirai. Mgmi. J.. 26: 259-276 (2005)
Marketing and Technology Resource Cotnplementarity 271
combinations do not always lead to synergistic per-formance impact and managers should avoid over-investing in contexts where resources cannot beleveraged through configuration, complementarityand/or integration. In terms of resource-based the-ory, synergistic rents cannot always be obtained.
Overall, the following picture emerges. In lowtechnologically turbulent environments, marketing-related capabilities (beta = 0.61) and technology-related capabilities (beta = 0.53) had similar maineffects and there was no interaction. In high tech-nologically turbulent environments, the techno-logy-related capabilities —3- performance beta(0.59) was greater than the marketing-related capa-bilities —> performance beta (0.29), but in additionthere was a significant interaction effect (beta =0.10). The main effect of marketing-related capa-bilities on performance appeared to decrease as theenvironment becomes more technologically tur-bulent, while (1) the effect of technology-relatedcapabilities remained unchanged and (2) the inter-action effect increased. However, it should benoted that when an interaction effect is signifi-cant the path coefficients represent the conditionaleffects of one capability when the other capahilityis at its mean. Thus, in high turbulence, the impactof marketing-related capabilities on performanceincreased with the level of technology-related capa-hilities and the impact of technology-related capa-bilities on performance increased with the level oftnarkeiing-related capahilities.
Overall, for high technologically turbulent envi-ronments, our results showed that the more thecapability in one area (i.e., marketing-related ortechnology-related), the higher the impact on per-formance of one more unit of the other capa-bility. Searching for such synergies and extract-ing synergistic rents is, of course, an importantmanagerial concem. But it is also an impor-tant theoretical concem in resource-based theory,which has long claimed the possibility of synergythrough complementarities. Our research demon-strates empirically such synergy for JVs operat-ing in high technologically turbulent environments.The results also support the dynamic capabilitiesview's contention that in high-velocity marketsthe outcomes of dynamic capabilities are particu-larly unpredictable (Eisenhardt and Martin, 2()(X)).This unpredictability may be attributable to theinteraction effect being significant only in thehigh turbulent environment. Future reseaich shoulddetermine whether other capabilities have similar
performance impact profiles (i.e., characterized bysynergistic interaction) and under what environ-mental conditions.
CONCLUSION
The value of our analyses is to show that resources(i.e., marketing-related capabilities and technol-ogy-related capabilities) and combinations ofresources (i.e., the interaction of capabilities) pro-duce different performance results when the con-text varies (i.e., high vs. low technologically tur-bulent). Often researchers posit linear main effectswith no interactions for independent, orthogonalvariables under a broad scope of conditions. How-ever valid as a first approximation, the loss ofrealism is severe. At times, the results will bevery misleading and managers who implementaccordingly will have counter-productive perfor-mance results. In this study, complex conditions(i.e., moderation) mid non-independent effects ofexogenous, yet controllable, fimi inputs are mod-eled. In addition; (I) three control variables wereincorporated for their possible impact on the corerelationships; and (2) performance was measuredrelative to objectives, which means that a ptioticapabilities are factored in. Both of these charac-teristics of the analysis procedure serve to ensurerigorous testing of the hypotheses. This realismcomes at the price of a more complex computa-tional load, yet simple but powerful insights areavailable to managers as a result. Lack of con-textual variation often leads to results so generalthat the conclusions are meaningless for manage-rial puiposes and misleading for theory testing anddevelopment purposes.
REFERENCES
Achrol RS. 1991. Evolution of the marketing organi/a-tion: new fonns for lurbulcnl environments. Journalof Marketing 55: 77-93.
Bagozzi RP. Edwards JR. 1998. A general approachTor representing constructs in organi/ational research.Organizaliiinat Research Methods 1: 43-87.
Barney J. 1991. Firm resources and sustained competitiveadvantage. Jounuil of MaruigemetU 17: 99-120.
Bollen KA. 1989. Structural Equations with UitcnlVariables. Wiley: New York.
Buckiin LP, Sengupia S. 1993. Organizing successful co-marketing alliances. Journal of Marketing 57: 32-46.
Clark KB, Fujimoto T. 1991. Product Development Per-formance: Strategy, Organization, and Management in
Copyright © 2005 John Wiley & Sons. Lid. Sin/r. t. J.. 26; 2f'9-ll(> (2(105)
272 M. Song et al.
the World Auto Industry. Harvard Business SchoolPress: Boston. MA.
Colli.s DJ. 1991. A resource-based analysis of globalcompetiiion: ihc case ot ihe bearings industry.Strarefiic Managemeni JotirnuL Summer Special Issue12: 49-68.
Das TK, Teng BS. 2n(K). A resource-based theory ofslniicgic dWianccs. Journal of Mamtfiemenr 20: 31-61.
Day G. 1990. Market Driven Siratef^y: Processes forCreatini' Value. Free Press: New York.
Day GS. 1994. The capabililies of market-drivenorganizalions. .hmrtial of Marketing 58: 37-52.
Day GS. Wensley R. 1988. Assessing advantage: aframework for diagnosing competitive superiority.Journal (if Marketing 52: 1-20.
Dillman DA. 1978. Mail and Tetephotte Surxeys: TheTotal Design Method. Wiley: New York.
Downey HK. Slwum JW. 1975. Environmental uncer-tainty: the construct and its applications. Academy ofManagetm-nt Jottrnal 18: 562-577.
Do/ YL. Hamel G. 1998. Attiattce Advantage: The Art ofCreating Value through Partnering. Harvard Bu.sinessSchool Press: Boston. MA.
Fiisenhardt KM. Martin JA. 200(). Dynamic capabilities:what are Ihey'l Strategic Management Journat. SpecialIssue 21(10-11): I I05-I121.
Emerson RM. 1962. Power-dependence relations. Amfr-ican Sociological Review 27: 31-41.
Gaski JP, 1984. The theor\' of power and conflict Inchannel ol" distribution. Journal of Marketing 48:9-29.
Grant RM. 1991. The resource-based theory of compet-itive advantage: implications for strategy formation.California Mamigemetit Review 33: 114-135.
Harrison JS. Hiti MA. Hoskisson RE. Ireland RD. 2(K1I.Re.source compiementiirity in business combinations:extending the logic to organizational alliances. Journalof Management 27: 679-690.
Hayduk LA. 1987. Structural Equation Modeling withLiSHEL: Essentials and Advances. John HopkinsPress: Baltimore. MD.
Heide J. 1994. hiterorganizaiional governance in market-ing channels. Journal of Marketing 58: 71-85.
Hellat CE. 20<X). Guest editor's introduction to thespecial issue: the evolution of firm capacities. StrategicManagemeni Jimrnal. Special Issue 21( 10-11):955-959.
Jaccard J, Wan CK. 1995. Measurement error in theanalysis of interaction effects between continuouspredictors using multiple regression: multiple indicatorand stmctural equation approaches. PsychologicalBulletin 117: 34S-.357.
.laccard J. Wan CK. 1996. LISREL Approaches toInteraction Effects in Multiple Regression. Sage:Thousand Oaks, CA.
Jaccard J, Turrisi R, Wan CK. 1990. Ititeraction Effectsin Multiple Regressitm. Sage: Newbury Park. CA.
Joreskog KG, Yang F. 1996. Nonlinear StructuralEquation Models: The Kenny-Judd Model withInteraction Effects. Erlbaum: Mahwah, NJ.
Kendall MG. Straut A. 1958. The Advanced Theory ofStati.\tics. Griffin: London.
Kenny DA. Judd CM. 1984. Estimating the nonlinearand interaction effects of latent variables. AmericanPsychology As.'iociations 96: 201-210.
Kogut B. 1988. Joint venture: theoretical and empiricalperspectives. Strategic Matiagetnetit Journal 9(4):319-332.
Kogul B. Zander U. 1992. Knowledge of the firm,ctimbinaiive capabililies, and the replication oftechnology. Organization Science 3: 338-397.
Kotabe M. Swan KS. 1995. The role of strategic ullianeesin high-technology new product development. Strate-gic Managetnetu Journal 16(8): 621-636.
Leonard-Barton D. 1992. Core capabilities and corerigidities: a paradox in managing new product devel-opment. Strategic Management Joutnat, Summer Spe-cial Issue 13: It 1-125.
Lieberman M. Montgomery DB. 1988. Firsi-moveradvantages. Strategic Management Journal, SummerSpecial Issue 9: 41-58.
LIppman S. Rumelt RP, 1982. Uncertain irrilability: ananalysis of inierlimi differences in effieiency undercompetition. Belt Journal of Economics 13: 418-438.
Madhok A. 1997. Cost, value and foreign market entrymode: the transaction and limi. Strategic ManagetnentJournal 18(1): 39-61.
Milliken FJ. 1987. Three types of perceived uncertaintyabout the environment: state, effect, and respon.seuncertainty. Acadetnv of Management Review 12:133-143.
Mint/berg H. 1987. Crafting strategy. Harvatd BusinessReview 65(1): 66-75.
Mowery DC. Oxlcy JE, Silverman BS. 1996. Strategicalliance and intedinn knowledge transfer. StrategicManagement Journal. Winter Special Issue 17:77-91.
Murray AL Siehl C. 1989. Join! Venture and OtherAlliatices: Creating a Succes.sful Cooperative Link-age. Financial Executive Research Foundation: Mor-ri.stown. NJ.
Noda T, Collis DJ. 2001. The evolution of intraindustryHrm heterogeneity: insights from a process study.Academy of Managemeni Journal 44: 897-925.
Page AL. 1993. Assessing new product developmentpractices and pertbrmance: establishing crucial norm.Journal of Product Innovation Management 10:273-290.
Park SH. Ungson GR. 1997. The effect of national cul-ture, organizational complementarity, and economicmotivation on the joint venture dissolution. Acadetnyof Management Journal 40: 279-307.
Penrose ET. 1959. The Theory of the Growth of the Firm.Wiley: New York.
Peteraf MA, 1993. The cornerstones of competitiveadvantage: a resource-based view. Strategic Manage-ment Journal 14(3): 179-191.
Pfeffer J. Salancik GR. 1978. The External Conrmt ofOrgan izatiotis: A Resource Dependence Perspective.Harper & Row: New York,
Ping RAJ. 1995. A parsimonious estimating technique forinteraction and quadratic latent variables. Journal ofMarketing Re.search 32: 336-347.
Ctipyrighl © im5 John Wilty & Son?!. Liii. Stral. Mgmi. J.. M: 259-276 (2005)
Marketing and Technology Resource Complementarity 273
Pisano GP. 1994. Knowledge, integration, and thelocus of learning: an empirical analysis of processtievelopment. Strategic Management Joumal, WinterSpecial Issue 15: S5-1(X).
Priihiilail C'K. Hamel G, 1990. The core competence ofthe corporation. HatrarJ Business Review 68: 79-91.
Song M, Montoya-Weiss MM. 2(K)L The effect ofperceived technical uncertainty on Japanese newproduct deveiopnienl. Academy of ManagementJoumal 44: 61-80.
Tan JJ. Litschert RJ. 1994. Environment-strategy rela-tionship and its performance implications: an empiri-ciil study of the Chinese electronics industry. StrategicManagetnent Journal 15(1): 1-20.
Teece DJ. Pisano G. Shucn A. 1997. Dynamic capabili-ties and strategic management. Strategic ManagementJournal 18(7): 509-,'i33.
Tushman ML, Andereon P. 1986. Technological discon-tiniMties and organizational environments. Administra-tive Science Quarterly 31: 4.19-465.
Wernerfcli B. 1984. A resource-based view of Ihe lirm.Strategic Management Journal 5(2): 171-180.
Wind J. Mahajan V. 1997. Issues and oppttnunities innew product development: an introduction lo thespecial issue. Joumat of Marketing Research 34:1-12.
WotKlccK-k CP. Beamish PW. Makino S. 1994.Ownership-based entry mode strategies and interna-tional performance. Journal of tnte mat tonal BusinessStudies 25: 253-273.
Xie J. Sirbu M. 1995. Price competition and compatibil-ity in the presence of positive demand externalities.Management Sciences 41: 909-926.
Copyrighi © 2005 John Wiley & Soni. Ltd. Strat. Mgmt. J.. 2Jb. 259-276 (200S)
274 M. Song et al.
= Q —c a d
^ ' ^ ^ Q.
•^ 5 1 - ^— " o >.w = — oQ S 5 I
cod
—d1
rNIN (N
dI
<N
o
1
-
DO1
oo1
o--)
3o
oo1
(N
o1
3'-'
oo1
cO
3o
i n
o1
go1
so1
INoo
inor>
o
T
£o1
rioo1
so
o
- l ;
c1
o
1
cc1
f l
o
3o
o
-005
o1
I-"!
oo
o
1^ r^ CJo o o
I
•g- r-- r-lo o oo c o
f , r oo c So o o1
—c f l JJ
o o oo o o
SS3o o o
1 1
S c So o o
o1
o1
=o1
o
oo
s
o
co1
sT
so
:
o1T
m 00
o1
-
o1
oI
:
o
o1
t
o1
:
: :o
2
o1
^
o1
:
r-
o1
:^
o1
rio
1
-
-0.
r--
O1
-0
o
;
mo1
o1
o1
i n
-0
£
o1
o
o1
-0
-o1
f N0-
3 oo(^rj on- t^oood d d d d d d d d
doTd O d d
od
od d
0 0 o*r-J
do*d
•n ; in: 00
d
og go cy r-t : * : ? - —q q q q o c —o d d d d d d
qpqpqqqqdddddddd
q q
f Nr id d d d
1 / ^
o — d q q — q qdddddddd
I I Irj^aorj^oooq q c — o o o —ddddddddI I I I° P Pd d d d d dI I I I
=! OI I
o oI
o o o oI I I
'if— • o
;
64
o
'1 (N
I
o
r i
o
12f N
O
00
oT• n
O1
i r ir j
3c
00
oo1
r-
o1
fNr j
r f
C
07
o1
—r i
o
rito
'Clo
d
•<*d
d
sO
rtrto
so (NOC Crr i n
3d1
i n
d
od
i n
3o
do
sc
•r.
oo
o
o
i n
0
o
go
o'c
•or:
moo
oo
-
o
i n
o
o
3o
i n
O
-
o1
e^i
1
,-«
O1
so1
08
o
?
i s S t S ' s ^
'= - E =
I '' §— = = =: s -=
S S co o o
^M^'E^t
II II
i|S I IIII: 5 1l
(d - 3 JO
1 g
•c s «; o o™ C ** '^ '^r "S " "S "s3
5 c
E "3 c u au .E T 3 ~ 2 J j>
Copyrighl ® 2005 John Wiley & Sons. Lid, Slrat. Mgmt. / . 26: 259-276 (2005)
Marketing and Technology Resource Complementarity 275
APPENDIX 2: SEM ANALYSIS OFINTERACTION EFFECTS
variance of an interaction latent construct is asfollows:
There have been several approaches lo rnterac-lion effect analysis using SEM. The Kenny andJudd (1984), Jaccard and Wan (1995), and Ping(1995) approaches arc based on the same assump-tions, namely that the linear latent constructs andmeasurement errors of their indicators are nor-mally distributed and have means of zero. ThePing (1995) approach, analogous to two-step esti-mation, can be seen as an approximation of theKenny-Judd method because the first siep is toanalyze the model without the interaction indica-tors. The parameters from this iirst analysis arethen used for calculating the parameters of theinteraclion variables, which are then specified asfixed parameters in tbe subsequent analyses. Thesecond approach involves adding additional latentconstructs to account for loadings and error vari-ances of the interaction indicators (Hayduk. 1987).This approach, however, can become impracticalwhen the mtxiel consists of several latent interac-tion constructs and multiplicative indicators. Thefinal approach is also based on the Kenny-Juddmodel, but it includes constant intercept terms inthe model (Joreskog and Yang. 1996). However.this latter approach requires elaborate and complexequations to specify the constraints.
The approach used in our study is in line withthat of Kenny and Judd (1984). Our approachinvolves first centering the raw scores and thussimplifying many of the mathematical relationsbetween variables and rendering the effects ofseveral of the constraints negligible (Jaccard andWan. 1996). Consequently, this method enables usto incorporate ai! the multiplicative indicators ofthe interaction effect into the model. The equationsof the measurement models of latent independentconstructs. TM (marketing-related capabilities) andFj (technology-related capabitities) are, in deviateform;
T, - Xr. Fr -H ej,
(1)
(2)
The first step in establishing the theoreticalconstraints for interaction effect estimation isto define the variance for the latent interactionconstruct (F^/FT). According to the previouslynoted assumptions (Kenny and Judd. 1984). the
= Var(FM)Var(Fr) M. FT)' (3)
Equation 3., specifying the variance for the latentinteraction construct, is constraint 01.
The second step is to establish the palh cwf-ficients between latent interaction construct andits multiplicative indicators (i.e.. X) as wellas to establish the error variances (i.e., cw^rj)for the indicators. From Equations 1 and 2,we see that MJ, = (X^HFM +eMi){XTjFT + fr>)= {XMXTJFMFT) + (Xx/.F^eTj) + {Xj,FTeMi) -hesui'T, = {XM,TJFMFT) + eM,r,- Therefore con-straint #2. defining the path coefficients (A)between latent interaction construct (FMF]) andits multiplicative indicators (MT), is:
With errors of the product indicators as:
j) + +
(4)
(5)
Therefore the residual variances of product indica-tors are:
+And since F^ and C/^. F/ and e^,,, as well ase^, and eTj, are each assumed to be uncorrelated.constraint #J. defining the residual variances ofproduct indicators, is:
(7)
Equation 6 shows that ihe residual variances ofproduct indicators, f.w.T/. is composed of threecompKinents. namely X'wiVar(F,w('7y). A-/^Var(FTe^i) and \ar(eM,eTj)- These compt)nents fonnthe basis of one set of non-linear equality con-straints, which necessitates constraint #3 (.Equation7). Constraint #3 also indicates that ffM,er> niayshare the same variance components among them-selves, which may result in correlated errors be-tween these eMieji- For example, e^iery and
Copyright ® 2005 John Wiley & Sonv. t_id. Stmi. Mgmt. J.. 26: 259-:!76 (2005)
276 M. Song cr al.
share Var(<'7,). Therefore all the pathsbetween error terms of niulliplicative indicatorsmust be freed e.xccpt where there is no variancesharing. For our analysis, the matrix below shows0 = Fixed Path (the two multiplicative error termsare not allowed to covary) and 1 = Free Path(the two multiplicative error lerms are allowed tocovary). This set of fixed paths establishes con-.Urainl ^4.
latent constructs. It has been shown that whenvariables are normally distributed and mean cen-tered, the covariance of each with the latentinteraction construct is zero (Kendall and Straut.1958: Kenny and Judd. 1984). This is because
("win 111100100
1:0i0010
1(101001
111100
11(110
1001
E(FT). Since all odd moments are zero, thenEiF^FjFr) equals 0 and since all mean cen-tered variables have expected values of zero, then£(^7-) also equals 0. Then E[COV{FMFT. FJ)] is0 and similarly £|Cov(F,v,Fr. FM)\ is 0. It fol-lows that the correlations between the marketing-related capabilities construct and the latent interac-tion construct, as well as between the technology-related capabilities construct and the latent inter-action construct, are zero. Therefore the fmal am-siraint #5 is:
'FM.FMFT = <t>FT.FMFT = 0 (8)
The Mnal step is to establish the covariancebetween the latent interaction construct and other
Copyrighl •£> 2005 Juhn Wiley & .Sons. Ltd. Simt. Mgmi. J.. 26: 259-276 (2005)
