《产业经济学》课程教学资源（文献资料）合并和收购：它们对高技术产业企业创新绩效的影响

Availableonlineatwww.sciencedirect.comresearoSCIENCIRECTpolicyELSEVIERResearch Policy 35 (2006) 642654www.elsevier.com/locate/respolMergers and acquisitions: Their effect on the innovativeperformance of companies in high-tech industriesMyriam Cloodt a,*, John Hagedoorn b,1, Hans Van Kranenburg c.2^ECIS and Organisation Science and Marketing (OSM),Departmentof Technology Management(TM),TUleTechnischeUniversiteitEindhoven,PO.Box513.5600MBEindhoven,TheNetherlandsbMERITand Department of Organization and Strategy,Faculty of Economics and Business Administration,Maastricht UniversityP.O.Box616,6200MDMaastricht,TheNetherlands Radboud University Nijmegen, Nijmegen School of Managemen, P0.Box 9108,6500 HK Njmegen,The NetherlandsReceived 6 October 2004; received in revised form 6 July 2005; accepted 22 February 2006Available online 2 May 2006AbstractThis study examines the post-M&A innovative performance of acquiring firms in four major high-tech sectors. Non-technologicalM&As appear to have a negative impact on the acquiring firm's post-M&A innovative performance. With respect to technologicalM&As,a large relative size of the acquired knowledge basereduces the innovative performance of the acquiring firm.The absolutesize of the acquired knowledge base only has a positive effect during the first couple of years after which the effect turns around andwe see a negative effect on the innovative performance of the acquiring firm.The relatedness between the acquired and acquiringfirms' knowledge bases has a curvilinear impact on the acquiring firm's innovative performance. This indicates that companiesshould target M&A'partners' that are neither too unrelated nor too similar in terms of their knowledge base2006ElsevierB.V.AllrightsreservedKeywords: M&As; Innovative performance; High-tech industries1.Introductiona preliminary explanation why M&As continue to beapopulargrowthstrategyof manycompanies(WorldContributions based on the resource-based view ofInvestmentReport,2000).In that context, itis stressedthe firm (Barney,1986,1991:Wernerfelt,1984),inthatopportunities fororganizational learning increasecombination with related work that stresses the impor-when a firm is exposed to new and diverse ideas basedtanceof organizational learningand innovation (Conneron differences in technological capabilities between theand Prahalad,1996;Grant,1996;Levitt and Marchacquiring and the acquired firm(see also Ghoshal,1987:1988:Nonaka,1991),provide someusefulinsights andHitt et al., 1996). Acquiring diverse external knowledgebases and making proper use of this newknowledge arefoundtoberelevantcontributions to a firm's post-M&ACorresponding author.Tel.:+31402475242;fax:+3140 2468054.innovativeperformanceE-mail addresses: m.m.a.h.cloodt@tm.tue.nl (M. Cloodt),Our current study is clearly linked to recent researchj.hagedoorn@os.unimaas.nl (J.Hagedoorn)that has already made some progress in analyzing criti-h.vankranenburg@fm.ru.nl (H.VanKranenburg)cal success factors that have a significant influence on a1 Tel.: +31 43 3883823; fax: +31 43 3884893.firm'spost-M&Ainnovativeperformance.Forinstance2Tel.: +31 243612028; fax: +3124 3611933.0048-7333/$ - see front matter 2006 Elsevier B.V. All rights reserved.doi:10.1016/jrespol.2006.02.007

Research Policy 35 (2006) 642–654 Mergers and acquisitions: Their effect on the innovative performance of companies in high-tech industries Myriam Cloodt a,∗, John Hagedoorn b,1, Hans Van Kranenburg c,2 a ECIS and Organisation Science and Marketing (OSM), Department of Technology Management (TM), TU/e Technische Universiteit Eindhoven, P.O. Box 513, 5600 MB Eindhoven, The Netherlands b MERIT and Department of Organization and Strategy, Faculty of Economics and Business Administration, Maastricht University, P.O. Box 616, 6200 MD Maastricht, The Netherlands c Radboud University Nijmegen, Nijmegen School of Management, P.O. Box 9108, 6500 HK Nijmegen, The Netherlands Received 6 October 2004; received in revised form 6 July 2005; accepted 22 February 2006 Available online 2 May 2006 Abstract This study examines the post-M&A innovative performance of acquiring firms in four major high-tech sectors. Non-technological M&As appear to have a negative impact on the acquiring firm’s post-M&A innovative performance. With respect to technological M&As, a large relative size of the acquired knowledge base reduces the innovative performance of the acquiring firm. The absolute size of the acquired knowledge base only has a positive effect during the first couple of years after which the effect turns around and we see a negative effect on the innovative performance of the acquiring firm. The relatedness between the acquired and acquiring firms’ knowledge bases has a curvilinear impact on the acquiring firm’s innovative performance. This indicates that companies should target M&A ‘partners’ that are neither too unrelated nor too similar in terms of their knowledge base. © 2006 Elsevier B.V. All rights reserved. Keywords: M&As; Innovative performance; High-tech industries 1. Introduction Contributions based on the resource-based view of the firm (Barney, 1986, 1991; Wernerfelt, 1984), in combination with related work that stresses the impor￾tance of organizational learning and innovation (Conner and Prahalad, 1996; Grant, 1996; Levitt and March, 1988; Nonaka, 1991), provide some useful insights and ∗ Corresponding author. Tel.: +31 40 2475242; fax: +31 40 2468054. E-mail addresses: m.m.a.h.cloodt@tm.tue.nl (M. Cloodt), j.hagedoorn@os.unimaas.nl (J. Hagedoorn), h.vankranenburg@fm.ru.nl (H. Van Kranenburg). 1 Tel.: +31 43 3883823; fax: +31 43 3884893. 2 Tel.: +31 24 3612028; fax: +31 24 3611933. a preliminary explanation why M&As continue to be a popular growth strategy of many companies (World Investment Report, 2000). In that context, it is stressed that opportunities for organizational learning increase when a firm is exposed to new and diverse ideas based on differences in technological capabilities between the acquiring and the acquired firm (see also Ghoshal, 1987; Hitt et al., 1996). Acquiring diverse external knowledge bases and making proper use of this new knowledge are found to be relevant contributions to a firm’s post-M&A innovative performance. Our current study is clearly linked to recent research that has already made some progress in analyzing criti￾cal success factors that have a significant influence on a firm’s post-M&A innovative performance. For instance, 0048-7333/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.respol.2006.02.007

643M. Cloodt et al. / Research Policy 35 (2006) 642-654Ahuja and Katila (2001) studied the impact of the abso-lar knowledge sources (Bierly and Chakrabarti, 1996)luteand relative sizeof acquiredknowledgebaseson theTherefore, it is the firm's ability to acquire, transferinnovativeperformanceoffirmsinthechemicalsindus-and integrate the acquired firm's knowledge base intotry.Our study is an extended replication of the analysistheknowledge base of the acquiring firm that creates abyAhuja and Katila (2001).Contraryto the applied sci-sustainablecompetitiveadvantage(Barney,1986).How-ences,replication studies arenot very popular in mostofever, we do realize that not all acquisitions are under-thesocialsciences,withthepossibleexceptionofappliedtaken for technological reasons with the soleintent toeconometrics and some fields in psychology.Hubbardlearn(Hamel,1991).M&AsmightalsobemotivatedandVetter(1992.1997)andFuess(1996)foundthatby market-entry and market-structure related consider-leadingjournalsineconomics,managementandfinanceations,or bythedesireto expand the firm'sproductpublish relativelyfew extended replication studies.Fromrangeinternationally(BerkovitchandNarayanan,1993Chakrabarti et al.,1994; Hagedoorn and Sadowski.a purely methodological point of view, this comes asquite surprisingbecause so far themanagement litera-1999:Trautwein,1990).Theseconsiderationsmotivateturehas generated a slightlyconfusing body of literature,firms to undertake non-technological acquisitions thatto say the least.For many basic questions in the litera-areless likelytoprovidetechnological knowledgetotheture,the reader will find contradictory findings,differentacquiring firm.3If M&As involvenooronlyafewtechnological com-measurements,unclearinternationalimplications.andthe use of a range of partlyoverlapping constructs.ponents, they are expected to have little or no effectTheAhujaandKatila(2001)studyconcentrated onon the innovation routines of the acquiring firm.How-the effect of M&As in a single medium-tech industrialever,if M&As create a disruption of the establishedcontext: the chemicals sector.Their contribution invitesroutines, thereby consuming significant managerial timesubsequent researchby others to consider a wider rangeand energy, they can havea negative impact on the post-of industries,in particular high-tech industries.OurM&Ainnovativeperformance(Ahuja andKatila,2001;HaspeslaghandJemison,1991;Hittetal.,1996).Instudy extends their work by analyzing the post-M&Ainnovativeperformancebymeans of a large sampleofparticular,acquisitions motivated by non-technologicalfirms operating infourhigh-tech sectors (OECD,1997):incentives,suchasshort-termprofitgrowth.canrequireaerospace and defense,computers and officemachin-so much managerial attention that this leads to a lowerery,pharmaceuticals,andelectronicsandcommunicamanagerial commitmenttolong-term investmentsininnovation(Hittetal.,1996).In summary,weexpecttions.Thesehigh-tech sectors are selectedfor twomainthat non-technological M&As either contribute little toreasons.First,these industries areprimarilyknowledgedriven industries (OECD, 1997).Technological learningthe innovative output of the acquiring firm, or that thereis expected tobe akey determinant in creating and sus-might be a negative impacton the post-M&Ainnovativetaining acompetitive advantagefor manyof the sampleperformance.Hence,firms in these industries (Bierly and Chakrabarti, 1996)Hypothesis 1.Non-technological acquisitions willSecondfor each of these industries.we can measurehave either a negative or a non-significant effect on theinnovativeperformance through the same indicator,ie.post-M&A innovative performance ofthe acquiringfirm.patents. It is well known that particularly in these indus-tries,patents playa significantrole in indicating impor-Thepossiblepositive impactoftechnological M&Astantaspects of innovative performance (Hagedoorn andoninnovativeperformancedependsonanumberoffac-Cloodt,2003;OECD,1997).Aswillbedemonstratedtors. A first critical dimension in the technological uni-below.ourextendedreplicationstudyisabletoconfirmfication of two firms concernsthe size of theacquiredsome findings of the Ahuja and Katila (2001) study butknowledgebases(Ahuja andKatila,2001).Theeffectitalsogeneratessomeimportantnewinsightrelatedtothe specific role of knowledge depreciation and time-constrained knowledge transfer through M&As in a3 As indicated by one of the referees, we focus in our paper on scalenumber of high-tech industries.and scope effects intechnological knowledgebutfirms are of coursealso concerned with scale and scope effects in products and industries.If we envisage the role of the firm as transforming technologies into2.Theoryand hypothesesproducts,non-technological acquisitions can expand sales and marketshares of products having a positive effect ontheeconomicperfor-According to theresource-based theory of the firmmance ofthe acquiringfirm.The same argument holdsfor undertakingand theknowledge-based view,differences in innova-a very closely related technological acquisition that may raise markettive performance between firms are a result of dissimi-power and patent exploitation

M. Cloodt et al. / Research Policy 35 (2006) 642–654 643 Ahuja and Katila (2001) studied the impact of the abso￾lute and relative size of acquired knowledge bases on the innovative performance of firms in the chemicals indus￾try. Our study is an extended replication of the analysis by Ahuja and Katila (2001). Contrary to the applied sci￾ences, replication studies are not very popular in most of the social sciences, with the possible exception of applied econometrics and some fields in psychology. Hubbard and Vetter (1992, 1997) and Fuess (1996) found that leading journals in economics, management and finance publish relatively few extended replication studies. From a purely methodological point of view, this comes as quite surprising because so far the management litera￾ture has generated a slightly confusing body of literature, to say the least. For many basic questions in the litera￾ture, the reader will find contradictory findings, different measurements, unclear international implications, and the use of a range of partly overlapping constructs. The Ahuja and Katila (2001) study concentrated on the effect of M&As in a single medium-tech industrial context: the chemicals sector. Their contribution invites subsequent research by others to consider a wider range of industries, in particular high-tech industries. Our study extends their work by analyzing the post-M&A innovative performance by means of a large sample of firms operating in four high-tech sectors (OECD, 1997): aerospace and defense, computers and office machin￾ery, pharmaceuticals, and electronics and communica￾tions. These high-tech sectors are selected for two main reasons. First, these industries are primarily knowledge￾driven industries (OECD, 1997). Technological learning is expected to be a key determinant in creating and sus￾taining a competitive advantage for many of the sample firms in these industries (Bierly and Chakrabarti, 1996). Second, for each of these industries, we can measure innovative performance through the same indicator, i.e. patents. It is well known that particularly in these indus￾tries, patents play a significant role in indicating impor￾tant aspects of innovative performance (Hagedoorn and Cloodt, 2003; OECD, 1997). As will be demonstrated below, our extended replication study is able to confirm some findings of the Ahuja and Katila (2001) study but it also generates some important new insight related to the specific role of knowledge depreciation and time￾constrained knowledge transfer through M&As in a number of high-tech industries. 2. Theory and hypotheses According to the resource-based theory of the firm and the knowledge-based view, differences in innova￾tive performance between firms are a result of dissimi￾lar knowledge sources (Bierly and Chakrabarti, 1996). Therefore, it is the firm’s ability to acquire, transfer and integrate the acquired firm’s knowledge base into the knowledge base of the acquiring firm that creates a sustainable competitive advantage (Barney, 1986). How￾ever, we do realize that not all acquisitions are under￾taken for technological reasons with the sole intent to learn (Hamel, 1991). M&As might also be motivated by market-entry and market-structure related consider￾ations, or by the desire to expand the firm’s product range internationally (Berkovitch and Narayanan, 1993; Chakrabarti et al., 1994; Hagedoorn and Sadowski, 1999; Trautwein, 1990). These considerations motivate firms to undertake non-technological acquisitions that are less likely to provide technological knowledge to the acquiring firm.3 If M&As involve no or only a few technological com￾ponents, they are expected to have little or no effect on the innovation routines of the acquiring firm. How￾ever, if M&As create a disruption of the established routines, thereby consuming significant managerial time and energy, they can have a negative impact on the post￾M&A innovative performance (Ahuja and Katila, 2001; Haspeslagh and Jemison, 1991; Hitt et al., 1996). In particular, acquisitions motivated by non-technological incentives, such as short-term profit growth, can require so much managerial attention that this leads to a lower managerial commitment to long-term investments in innovation (Hitt et al., 1996). In summary, we expect that non-technological M&As either contribute little to the innovative output of the acquiring firm, or that there might be a negative impact on the post-M&A innovative performance. Hence, Hypothesis 1. Non-technological acquisitions will have either a negative or a non-significant effect on the post-M&A innovative performance of the acquiring firm. The possible positive impact of technological M&As on innovative performance depends on a number of fac￾tors. A first critical dimension in the technological uni- fication of two firms concerns the size of the acquired knowledge bases (Ahuja and Katila, 2001). The effect 3 As indicated by one of the referees, we focus in our paper on scale and scope effects in technological knowledge but firms are of course also concerned with scale and scope effects in products and industries. If we envisage the role of the firm as transforming technologies into products, non-technological acquisitions can expand sales and market shares of products having a positive effect on the economic perfor￾mance of the acquiring firm. The same argument holds for undertaking a very closely related technological acquisition that may raise market power and patent exploitation

644M.Cloodt et al. / Research Policy 35 (2006)642-654of M&As depends on whether targets have a similar orThe integration of a knowledge base that is of a rela-preferablylargerR&Dinput and otherinnovative activi-tivelylargesizecandisruptexistinginnovativeactivitiesties.The unification of twoknowledgebases can provideand render the different integration stages more com-opportunities for synergies in future R&D, while reduc-plex,moretimeconsuming andfull ofrisks (Capron anding redundant or duplicate R&D efforts and provide aMitchell,2000;Chakrabartietal.,1994;HaspeslaghandJemison,1991).Duetosuchproblems,integratingarela-largerresearchbaseto financecosts (Cassiman et al.,2005:Hall, 1990).tivelylargeknowledgebaserequires additional resourcesAnother positive effect of the increased size ofto be devoted to integration activities, leaving fewerknowledge bases, is found in the potential for aggre-resourcesfor theactual innovative endeavor (Ahujaandgation (Grant, 1996).The transfer ofknowledge fromKatila, 2001).Thus,we expect that with the integrationthe acquired firm to the acquiring firm involves bothof a relatively large knowledge base, fewer resourcestransmission and receipt (Grant, 1996). Receipt canwill be available for innovative activities, which has abe analyzed in terms of a firm's absorptive capacity.negative impacton theacquirer's post-M&A innovativewhich plays a dual role in improving innovative per-performance.formance (Cohen and Levinthal, 1989, 1990).When aHypothesis 3.There is a negative relationship betweenfirm increases its internal knowledge base by acquiringknowledge,it can usethisknowledgeto generatenewthe relative size of the acquired knowledge base and theinnovations.In addition,the expansion of theinternalpost-M&Ainnovativeperformanceoftheacquiringfirm.knowledge base also increases thefirm's ability to rec-ognize thevalueof new information,to assimilate it andAthird importantfactorin themerger of two firmsto exploit itforcommercial ends(Cohen and Levinthal,is their relatedness in terms of particular fields of tech-1989).nology that the acquiring firm shares with the acquiredHence,byundertakingM&As,firmsarenotonly con-firm(Cassiman et al.,2005;Hagedoorn and Duysters,fronted with the internally created knowledge base of2002).WhiletherelatednessofM&Asintermsofthe acquired firm. By taking over the acquired firm'sproduct-markets concerns the industry-aspect,the tech-knowledgebase,thefirmwillbeableto view somenologicalrelatednessrefersto firm-specific aspects suchissues from a different perspective and recognize theas technological disciplines and engineering capabili-value of new external knowledge,which can help theties.The positive effect of relatedness in technologicalknowledge on the success of M&As is found by severalacquirer develop a richer knowledge base (Ahuja andKatila,2001;Levinthal and March,1993:VermeulenstudiesthatemphasizetheeffectsofeconomiesofscaleandBarkema,2001).Several studies mention the advanandscopeofR&D,suchasashorterinnovationlead-timetages of creating a richer or broader knowledge base,and the possibility to engage in larger combined projectssuch as increased strategic flexibility, sustainable com+(Gerpott,1995:HagedoornandDuysters,2002)petitive advantage, and increased performance (BierlyFrom an organizational learning perspective,this pos-andChakrabarti,1996;HendersonandCockburn,1994;itive effect lies in theabilityto better evaluateand utilizeReed andDeFillippi, 1990).We expectthatthe acquisi-related externally acquiredknowledge than unrelatedtion ofexternally availableknowledgeleads to increasedexternally acquired knowledge (Cohen and Levinthal,economies of scale and scope and a broaderknowledge1990).This is based on the idea that a firm's absorptivebase, both having a positive effect on innovative perfor-capacity depends mainly on its level of knowledge in aspecific field (Cohen and Levinthal,1990; Duysters andmance.Hagedoorn,2000:Moweryet al.,1996).Iftheknowl-Hypothesis 2.There is a positive relationship betweenedge base of the acquirer is not sufficiently adapted tothe absolute size of the acquired knowledge base andtheacquiredknowledge,theabsorptionprocessbecomesvery difficult (Duysters and Hagedoorn, 2000). There-thepost-M&A innovativeperformance of theacquiringfirm.fore, we argue that unrelated technologies often requirearadical change inthe way oforganizing research(KogutThe challengefor companies is not just to acquireand Zander,1992)which can easilybecounterproduc-knowledgebases butalso tointegratethem in ordertotive(Ahuja andKatila,2001;Dosi,1988).improvethepost-M&Ainnovativeperformance(AhujaHowever,technological knowledge and engineeringand Katila,2001:Child et al.,2001:Haspeslagh andcapabilities that are too similar to the already existingJemison,1991).This integration processforms the sec-knowledge of the acquiring company will contributelittle to the post-M&A innovative performance. Someond critical dimension in the unification of two firms

644 M. Cloodt et al. / Research Policy 35 (2006) 642–654 of M&As depends on whether targets have a similar or preferably larger R&D input and other innovative activi￾ties. The unification of two knowledge bases can provide opportunities for synergies in future R&D, while reduc￾ing redundant or duplicate R&D efforts and provide a larger research base to finance costs (Cassiman et al., 2005; Hall, 1990). Another positive effect of the increased size of knowledge bases, is found in the potential for aggre￾gation (Grant, 1996). The transfer of knowledge from the acquired firm to the acquiring firm involves both transmission and receipt (Grant, 1996). Receipt can be analyzed in terms of a firm’s absorptive capacity, which plays a dual role in improving innovative per￾formance (Cohen and Levinthal, 1989, 1990). When a firm increases its internal knowledge base by acquiring knowledge, it can use this knowledge to generate new innovations. In addition, the expansion of the internal knowledge base also increases the firm’s ability to rec￾ognize the value of new information, to assimilate it and to exploit it for commercial ends (Cohen and Levinthal, 1989). Hence, by undertaking M&As, firms are not only con￾fronted with the internally created knowledge base of the acquired firm. By taking over the acquired firm’s knowledge base, the firm will be able to view some issues from a different perspective and recognize the value of new external knowledge, which can help the acquirer develop a richer knowledge base (Ahuja and Katila, 2001; Levinthal and March, 1993; Vermeulen and Barkema, 2001). Several studies mention the advan￾tages of creating a richer or broader knowledge base, such as increased strategic flexibility, sustainable com￾petitive advantage, and increased performance (Bierly and Chakrabarti, 1996; Henderson and Cockburn, 1994; Reed and DeFillippi, 1990). We expect that the acquisi￾tion of externally available knowledge leads to increased economies of scale and scope and a broader knowledge base, both having a positive effect on innovative perfor￾mance. Hypothesis 2. There is a positive relationship between the absolute size of the acquired knowledge base and the post-M&A innovative performance of the acquiring firm. The challenge for companies is not just to acquire knowledge bases but also to integrate them in order to improve the post-M&A innovative performance (Ahuja and Katila, 2001; Child et al., 2001; Haspeslagh and Jemison, 1991). This integration process forms the sec￾ond critical dimension in the unification of two firms. The integration of a knowledge base that is of a rela￾tively large size can disrupt existing innovative activities and render the different integration stages more com￾plex, more time consuming and full of risks (Capron and Mitchell, 2000; Chakrabarti et al., 1994; Haspeslagh and Jemison, 1991). Due to such problems, integrating a rela￾tively large knowledge base requires additional resources to be devoted to integration activities, leaving fewer resources for the actual innovative endeavor (Ahuja and Katila, 2001). Thus, we expect that with the integration of a relatively large knowledge base, fewer resources will be available for innovative activities, which has a negative impact on the acquirer’s post-M&A innovative performance. Hypothesis 3. There is a negative relationship between the relative size of the acquired knowledge base and the post-M&A innovative performance of the acquiring firm. A third important factor in the merger of two firms is their relatedness in terms of particular fields of tech￾nology that the acquiring firm shares with the acquired firm (Cassiman et al., 2005; Hagedoorn and Duysters, 2002). While the relatedness of M&As in terms of product–markets concerns the industry-aspect, the tech￾nological relatedness refers to firm-specific aspects such as technological disciplines and engineering capabili￾ties. The positive effect of relatedness in technological knowledge on the success of M&As is found by several studies that emphasize the effects of economies of scale and scope of R&D, such as a shorter innovation lead-time and the possibility to engage in larger combined projects (Gerpott, 1995; Hagedoorn and Duysters, 2002). From an organizational learning perspective, this pos￾itive effect lies in the ability to better evaluate and utilize related externally acquired knowledge than unrelated externally acquired knowledge (Cohen and Levinthal, 1990). This is based on the idea that a firm’s absorptive capacity depends mainly on its level of knowledge in a specific field (Cohen and Levinthal, 1990; Duysters and Hagedoorn, 2000; Mowery et al., 1996). If the knowl￾edge base of the acquirer is not sufficiently adapted to the acquired knowledge, the absorption process becomes very difficult (Duysters and Hagedoorn, 2000). There￾fore, we argue that unrelated technologies often require a radical change in the way of organizing research (Kogut and Zander, 1992) which can easily be counterproduc￾tive (Ahuja and Katila, 2001; Dosi, 1988). However, technological knowledge and engineering capabilities that are too similar to the already existing knowledge of the acquiring company will contribute little to the post-M&A innovative performance. Some

645M. Cloodt et al. / Research Policy 35 (2006) 642-654degree of differentiation in technological capabilitiesthe summed coefficients can also be computed with thisbetween thefirms may enrich the acquiring firm'sknowl-model specification (Gujarati,1988)To control for unobserved heterogeneity,we gatherededgebaseandcreateopportunitiesforlearning(Ghoshal1987:Hittetal..1996).Thisenrichmentof the acquiringpre-sampleinformation oftheunobserveddifferences inknowledge stocksbetweenthesamplefirms.Weincludefirm'sknowledgebase and aproper use oftheexternalknowledge are relevantcontributions to a firm's innova-unobserved heterogeneity as an additional covariate intiveperformance(Cohen and Levinthal,1989;Grilichesthe model (Xir-1). However, possible unobserved firm1990;Pakes and Griliches,1984).In other words,weeffects can lead to serial correlation among the residu-expect that one has to strive for moderate relatednessalsofobservationsfromthesamefirm.Toaddressthisbetweenknowledgebases.On theone hand,theacquiredissue of unobserved heterogeneity we used the general-knowledge has to show enough overlap to facilitate theized estimating equations (GEE)estimation procedureabsorptionprocess.Ontheotherhand,thecombinationto estimate all models.This procedure provides a directof knowledgebasesrequires enoughdiversitytomakeapproach to modeling longitudinal count data with seriala substantial contribution to the post-M&A innovativecorrelation (Liang and Zeger, 1986)performance.Hence,3.2. Sample and dataHypothesis 4.The technological relatedness of theThe hypotheses are tested on a relatively large inter-acquired knowledge base will be curvilinearly (inversenational sample of companies covering four high-techU-shaped)related tothepost-M&Ainnovativeperfor-industries:aerospace and defense (SIC-codes 372 andmance of the acquiring firm.376),computersand officemachinery(SIC-code357)pharmaceuticals(SIC-code 283)and electronics and3.Methodscommunications(SIC-code36).Our sample consistsof347companiesof which21(6.05%)operateinthe3.1. Modelaerospace and defense industry,76 (21.9%)arefoundincomputersandofficemachinery,77(22.19%)areThis study uses a panel dataset model that combinesactiveinpharmaceuticals,and 173(49.86%)operate intimeseriesandcross-sectionstoanalyzeourdataandtestheelectronicsandcommunicationssector.Thesamplethehypotheses.FollowingAhujaandKatila(2001),weconsists of256NorthAmericancompaniesand91com-specify the following random effects negative binomialpaniesfromotherregions(45fromEuropeand46fromregression model:Asia).Our samplecanbeclassifiedas abalanced panelPir = exp(Xit-1y + Ait-1β1 + Ait-2β2datasetmeaningthatnofirmsexitedtheindustryorwereacquired by others during theperiod of our analysis.In+Ait-3P3 + Ait-4β4)addition, there were no firms that entered the sample at awhere Pir is a non-negative integer-valued count vari-later period in time. All the firms included in the sampleable for post-M&A innovativeperformance,measuredhavethesame startingpoint.by the number of patents achieved by firm i in year t,Our sample is also diverse in terms of the distributionXit-1 the vector of control variables affecting Pit (e.g.ofthe sizeof companies.About18%ofthecompaniesfirm size,industry,nationality,cultural distance,tmeandin our sample are relatively small withless than 1000employees.Almost the same percentage of companiesunobserved heterogeneity), Ait-year j the lagged vectorcan be characterized as very large with more than50,000of theindependentvariablesfor year j=1-4,ythevec-tor of regression coefficients for the control variablesemployees.Morethanhalf ofthesample(64%)canbeand theβs are the vectors of regression coefficients forfound in intermediate size-classesIntotal,weidentified2429M&Aeventsforourthe jth period lagged independent variables. By includ-ing lagged effects we can subsequently test the effect ofsample firms in the period 1985-1994. These M&Aacquisitionsfor upto4yearsaftertheyeartheM&Awasevents refer to the merging of two more or less equaloriginallymade.Thetotal impactof anM&Aacrosstimecompanies,as well as to acquisitions where one com-can be analyzed by summing theregression coefficientspany obtains majority ownership over another com-pany. To distinguish between technological and non-on the distributed lags.Bycalculating t-statistics we cantest the hypothesis that the total impact of acquisitionstechnological M&As,weanalyzed ifthetargetfirmhadsummed across all years,is zero and check whether it isanypatentingactivityin the5yearsprecedingtheM&Astatistically significant (Greene, 1993).Thevariance for(seeAhujaandKatila,2001).Of thetotal amountof

M. Cloodt et al. / Research Policy 35 (2006) 642–654 645 degree of differentiation in technological capabilities between the firms may enrich the acquiring firm’s knowl￾edge base and create opportunities for learning (Ghoshal, 1987; Hitt et al., 1996). This enrichment of the acquiring firm’s knowledge base and a proper use of the external knowledge are relevant contributions to a firm’s innova￾tive performance (Cohen and Levinthal, 1989; Griliches, 1990; Pakes and Griliches, 1984). In other words, we expect that one has to strive for moderate relatedness between knowledge bases. On the one hand, the acquired knowledge has to show enough overlap to facilitate the absorption process. On the other hand, the combination of knowledge bases requires enough diversity to make a substantial contribution to the post-M&A innovative performance. Hence, Hypothesis 4. The technological relatedness of the acquired knowledge base will be curvilinearly (inverse U-shaped) related to the post-M&A innovative perfor￾mance of the acquiring firm. 3. Methods 3.1. Model This study uses a panel dataset model that combines time series and cross-sections to analyze our data and test the hypotheses. Following Ahuja and Katila (2001), we specify the following random effects negative binomial regression model: Pit = exp(Xit−1γ + Ait−1β1 + Ait−2β2 + Ait−3β3 + Ait−4β4) where Pit is a non-negative integer-valued count vari￾able for post-M&A innovative performance, measured by the number of patents achieved by firm i in year t, Xit−1 the vector of control variables affecting Pit (e.g. firm size, industry, nationality, cultural distance, time and unobserved heterogeneity), Ait−year j the lagged vector of the independent variables for year j = 1–4, γ the vec￾tor of regression coefficients for the control variables, and the βs are the vectors of regression coefficients for the jth period lagged independent variables. By includ￾ing lagged effects we can subsequently test the effect of acquisitions for up to 4 years after the year the M&A was originally made. The total impact of an M&A across time can be analyzed by summing the regression coefficients on the distributed lags. By calculating t-statistics we can test the hypothesis that the total impact of acquisitions, summed across all years, is zero and check whether it is statistically significant (Greene, 1993). The variance for the summed coefficients can also be computed with this model specification (Gujarati, 1988). To control for unobserved heterogeneity, we gathered pre-sample information of the unobserved differences in knowledge stocks between the sample firms. We include unobserved heterogeneity as an additional covariate in the model (Xit−1). However, possible unobserved firm effects can lead to serial correlation among the residu￾als of observations from the same firm. To address this issue of unobserved heterogeneity we used the general￾ized estimating equations (GEE) estimation procedure to estimate all models. This procedure provides a direct approach to modeling longitudinal count data with serial correlation (Liang and Zeger, 1986). 3.2. Sample and data The hypotheses are tested on a relatively large inter￾national sample of companies covering four high-tech industries: aerospace and defense (SIC-codes 372 and 376), computers and office machinery (SIC-code 357), pharmaceuticals (SIC-code 283) and electronics and communications (SIC-code 36). Our sample consists of 347 companies of which 21 (6.05%) operate in the aerospace and defense industry, 76 (21.9%) are found in computers and office machinery, 77 (22.19%) are active in pharmaceuticals, and 173 (49.86%) operate in the electronics and communications sector. The sample consists of 256 North American companies and 91 com￾panies from other regions (45 from Europe and 46 from Asia). Our sample can be classified as a balanced panel dataset meaning that no firms exited the industry or were acquired by others during the period of our analysis. In addition, there were no firms that entered the sample at a later period in time. All the firms included in the sample have the same starting point. Our sample is also diverse in terms of the distribution of the size of companies. About 18% of the companies in our sample are relatively small with less than 1000 employees. Almost the same percentage of companies can be characterized as very large with more than 50,000 employees. More than half of the sample (64%) can be found in intermediate size-classes. In total, we identified 2429 M&A events for our sample firms in the period 1985–1994. These M&A events refer to the merging of two more or less equal companies, as well as to acquisitions where one com￾pany obtains majority ownership over another com￾pany. To distinguish between technological and non￾technological M&As, we analyzed if the target firm had any patenting activity in the 5 years preceding the M&A (see Ahuja and Katila, 2001). Of the total amount of

646M.Cloodt et al. / Research Policy 35 (2006)642-654M&As.1148mettheabove-mentionedcriterion andtheynumber of the patents that its acquisitions had obtainedare classified as technological M&As. The remainingduringthepreceding5yearsbeforetheparticularM&A1281M&Asareclassified as non-technological M&As.event.Thesepatents werethencombined withthepatentsForthefirms inthe sample,weobtained annual patentthat were cited by these companies.Duplicates werecount data for the period1980-1994 and acquisition andabstracted from the list to ensure that a patent codefirm-specific data for the years 1980-1993.The finalappears only once.The acquired knowledge base waspanel for the regression analysis amounts to 7 years fromthen calculated as the number of patents (i.e.knowledge1989to 1995.elements)onthis list.Itiswell known thatthere isno ‘official'databaseRelative size of acquired knowledge base.This vari-with a world-wide, industry level list of all companiesable was measured by dividing the absolute size offrom which one can draw a random sample. Our samplethe acquired knowledge base by the absolute size ofis taken from the Securities Data databank, which con-the acquiring firm's knowledge base.The absolute sizetains information on theyearanM&A was established,of theacquiring firm'sknowledgebasewas calculatedusing the same procedure as the absolute size of thethe acquirer.thetarget.theparent acquirerand thepar-ent target firm. Industry information is provided in SICacquired firm's knowledge base.In very few cases, thecodes of the acquiree and acquirer. Acquiring firms areacquired knowledge base was larger than the acquiringselected based on the industry information provided infirm'sknowledge base.In these cases,we usedthelargerSIC-codes which should cover one of the four high-technumber as the denominator. As we are interested in theindustries as mentioned above.Additional informationrelativeproportionofthemerged firm'sresourcesthatareonsizeandR&Dexpendituresofcompanieswasidenti-likely to be occupied with integrative rather than inven-fied through otherdatasets such as Amadeus,Compustat,tive activity,a number greaterthan one is not meaningful.and Worldscope.Data on patents and patentcitations areTechnologically related and technologically unre-takenfromtheUSPatentandTrademarkOfficedatabaselatedM&As.Tomeasuretherelatedness of theacquired(USDepartmentof Commerce).knowledge base,wecomposed a list of patent codes thatappeared in boththe acquiredfirm'sknowledge baseandin the acquiring firm's knowledge base.Thesepatents3.3.Variablesweredividedbythe absolute sizeof acquired knowledgebaseThedependent variable,post-M&A innovativeper-All independent variables described in the above con-formance of the acquiring firm, is measured by thesistof four lagged versions.number of patents granted to each acquiring firm.4 Wemeasure patentsi as the numberof successful patent3.4.Controlvariablesapplications orpatentsgranted,forthe acquiring firmi in year t.The dependent variable is based on the num-We control fora number of possible additional effectsber of patents of the acquiring firm obtained during 1-4on thepost-M&Ainnovativeperformanceof theacquir-years after theM&A.ing firm.Previous research on the effect of cultural dis-Number of non-technologicalM&As.M&Asaretance on post-M&Aperformance suggests bothpositivereported as technological acquisitions if the acquiredandnegative effects,but thenegative effects seemtobefirmhadanypatenting activityduringthe5yearspreced-dominant (Datta,1991;Haspeslagh and Jemison,1991).ing the acquisition.M&As that did not meet the above-WeusetheKogut and Singh (1988)modified indexofmentioned criterion are considered as non-technologicalHofstedetocontrolfor international cultural differencesM&As. To distinguish non-technological acquisitionsbetween companies involved in M&As.When calculat-fromtechnologicalacquisitions,weanalyzedthepatentingthisvariablewehavetomakeacorrectionasazeroingactivityof theacquiredfirm in the5yearsprecedingfor an observation in a certain year can representboththe M&A event.a domesticM&A (no cultural distance)and no M&AAbsolute sizeof acquiredknowledgebase.Foreach(technological or non-technological)in that particularacquiring firm and foreach year, a list was madewith theyear.We use a dummy variable to correct for this by set-ting the values of a dummy variable cultural distance toone,each year no M&A took place.4 Contrary to Ahuja and Katila (2001), our dependent variable isStudies by Griliches (1998) and Pakes and Grilichesmeasured by patentsgranted and we do not assign a granted patent(1984)indicatea statistical relationshipbetweenR&Dto the year in which it was originally applied for due to right handcensoring and data limitations.and the number of patents,although patent output

646 M. Cloodt et al. / Research Policy 35 (2006) 642–654 M&As, 1148 met the above-mentioned criterion and they are classified as technological M&As. The remaining 1281 M&As are classified as non-technological M&As. For the firms in the sample, we obtained annual patent count data for the period 1980–1994 and acquisition and firm-specific data for the years 1980–1993. The final panel for the regression analysis amounts to 7 years from 1989 to 1995. It is well known that there is no ‘official’ database with a world-wide, industry level list of all companies from which one can draw a random sample. Our sample is taken from the Securities Data databank, which con￾tains information on the year an M&A was established, the acquirer, the target, the parent acquirer and the par￾ent target firm. Industry information is provided in SIC codes of the acquiree and acquirer. Acquiring firms are selected based on the industry information provided in SIC-codes which should cover one of the four high-tech industries as mentioned above. Additional information on size and R&D expenditures of companies was identi- fied through other datasets such as Amadeus, Compustat, and Worldscope. Data on patents and patent citations are taken from the US Patent and Trademark Office database (US Department of Commerce). 3.3. Variables The dependent variable, post-M&A innovative per￾formance of the acquiring firm, is measured by the number of patents granted to each acquiring firm.4 We measure patentsit as the number of successful patent applications or patents granted, for the acquiring firm i in year t. The dependent variable is based on the num￾ber of patents of the acquiring firm obtained during 1–4 years after the M&A. Number of non-technological M&As. M&As are reported as technological acquisitions if the acquired firm had any patenting activity during the 5 years preced￾ing the acquisition. M&As that did not meet the above￾mentioned criterion are considered as non-technological M&As. To distinguish non-technological acquisitions from technological acquisitions, we analyzed the patent￾ing activity of the acquired firm in the 5 years preceding the M&A event. Absolute size of acquired knowledge base. For each acquiring firm and for each year, a list was made with the 4 Contrary to Ahuja and Katila (2001), our dependent variable is measured by patents granted and we do not assign a granted patent to the year in which it was originally applied for due to right hand censoring and data limitations. number of the patents that its acquisitions had obtained during the preceding 5 years before the particular M&A event. These patents were then combined with the patents that were cited by these companies. Duplicates were abstracted from the list to ensure that a patent code appears only once. The acquired knowledge base was then calculated as the number of patents (i.e. knowledge elements) on this list. Relative size of acquired knowledge base. This vari￾able was measured by dividing the absolute size of the acquired knowledge base by the absolute size of the acquiring firm’s knowledge base. The absolute size of the acquiring firm’s knowledge base was calculated using the same procedure as the absolute size of the acquired firm’s knowledge base. In very few cases, the acquired knowledge base was larger than the acquiring firm’s knowledge base. In these cases, we used the larger number as the denominator. As we are interested in the relative proportion of the merged firm’s resources that are likely to be occupied with integrative rather than inven￾tive activity, a number greater than one is not meaningful. Technologically related and technologically unre￾lated M&As. To measure the relatedness of the acquired knowledge base, we composed a list of patent codes that appeared in both the acquired firm’s knowledge base and in the acquiring firm’s knowledge base. These patents were divided by the absolute size of acquired knowledge base. All independent variables described in the above con￾sist of four lagged versions. 3.4. Control variables We control for a number of possible additional effects on the post-M&A innovative performance of the acquir￾ing firm. Previous research on the effect of cultural dis￾tance on post-M&A performance suggests both positive and negative effects, but the negative effects seem to be dominant (Datta, 1991; Haspeslagh and Jemison, 1991). We use the Kogut and Singh (1988) modified index of Hofstede to control for international cultural differences between companies involved in M&As. When calculat￾ing this variable we have to make a correction as a zero for an observation in a certain year can represent both a domestic M&A (no cultural distance) and no M&A (technological or non-technological) in that particular year. We use a dummy variable to correct for this by set￾ting the values of a dummy variable cultural distance to one, each year no M&A took place. Studies by Griliches (1998) and Pakes and Griliches (1984) indicate a statistical relationship between R&D and the number of patents, although patent output

647M. Cloodt et al. / Research Policy 35 (2006) 642-654appears to gradually decrease with an increase in R&Dindependent variables, with the expected exception ofexpenditures.The control variable foryearly R&Drelatedness of the acquired knowledge base with itssquared term.There is also little correlation between theexpendituresisstandardizedbyconvertingthedatafromnational currenciesto US dollarsindependent variables and the control variables.OnlyPrevious studies indicate that the patent activity ofthe correlation between pre-samplepatents and R&Dcompanies increases with size(Cohen and Levin, 1989:is rather high (0.784),although lower than reported byMansfield,1986).The size of the acquiring firm is mea-Ahuja and Katila (0.89).Since we have to control forsured bytaking the natural log of its number of employ-unobserved heterogeneity and also because we prefer toees.Wealso control for differencesbetween the fourfollowtheAhuja andKatila(2001)model,wewillkeephigh-tech sectors.All four sectors are high-tech indus-both variables in ourmodel.tries but there is still considerable dissimilarity in theirTable1displays the estimation results ofthe negativeR&D and patenting-intensity (OECD, 1992),which canbinomial model with distributed lag analysis forthe postpartly beexplained by inter-sectoraldifferences betweenM&Ainnovativeperformanceofthesamplefirms.Wethenature of knowledge and the regime of appropri-use the full model to discuss our results.The results ofability.We included sector dummies to controlfor thesethe other models are presented in Appendix Bdifferences.Giventhe international differences inpatentHypothesis 1 argues that non-technological M&Asingbehaviorbetween companiesfrom theUSA,Europecontribute little to the innovative output of the acquir-andJapan(OECD,2001),wecontrolforthenational-ing firm or thattheremight evenbe a negative impactity of theacquiring firm.ANational ScienceFoundationon thepost-M&A innovative performance.The indi-(1998)study shows thatthe number of patents increasedvidual coefficients of thevariable for the number ofsignificantly during the period of our study.To controlnon-technological acquisitions are significant and threefor this we included year dummies.out of four show a negative sign.The summed coef-Pre-sample differences in thenumber of patents officient, reflecting the total impact, is negative andthe sample firms should have a positive influence on thesignificant. Together this provides strong support forinnovative performance of the acquiring firm after theHypothesis 1.M&A.Firms that achievehigh levels of innovativenessWe also anticipate that the acquisition of the targetfirm's knowledge base will lead to increased economieshave,according toDosi (1988),alsoa higher chance ofmaintaining or increasing their level of innovativenessof scaleand scopeandabroaderknowledgebase.BothThe firm heterogeneity control variable pre-sampleare expected tohave apositive effect on the post-M&Apatents is measured as the sum of patents obtained byinnovative performance (Hypothesis 2).The individuala firm in the 3years prior to the firm's entryinto thecoefficients of the variable absolute size of acquiredknowledgebaseare all significantbut show a diversesample.As already mentioned in the previous section, the rel-picture.The absolute size of acquired knowledge baseative size of the acquired knowledge base is calculatedimprovespost-M&Ainnovativeperformanceduringtheby dividing the absolute size of the acquired knowledgefirst 2 years, but thereafter it has a negative influ-basebythe size ofthe acquiring firm'sknowledgebaseence.The summed coefficient is significant and neg-However, in a few cases, the acquiring firm's knowl-ative, which also provides no support for Hypothesisedge base is zero for some of the years in the panel data2. Similar to the study of Ahuja and Katila, the totalset. In order to calculate the relative size correctly, weimpact of the absolute size of the acquired knowledgefollow Hausman et al. (1984) and set these specific val-base on post-M&A innovative performance is ratherues to one instead of zero and use a dummy variable.small given that a one-unit increase in the absolute sizeof the knowledge base leads to a 0.02% decrease incalled dummy absolute size acquirers, to correctfor thissolution.thepost-M&Ainnovativeperformanceof theacquiringfirm.Hypothesis3 predicts a negative relationship between4. Resultsthe relative size of the acquired knowledge base andthe post-M&A innovative performance of the acquir-Appendix A provides the descriptive statistics for allingfirm.The individual and summed coefficients arevariables.5There is little correlation between themainall significant and negative, which clearly supports thishypothesis.Hypothesis 4 states that the acquired knowledge base We do not report the extensive correlation statistics,this informa-has to show enough overlap with the acquirer's alreadytion is available from the authors

M. Cloodt et al. / Research Policy 35 (2006) 642–654 647 appears to gradually decrease with an increase in R&D expenditures. The control variable for yearly R&D expenditures is standardized by converting the data from national currencies to US dollars. Previous studies indicate that the patent activity of companies increases with size (Cohen and Levin, 1989; Mansfield, 1986). The size of the acquiring firm is mea￾sured by taking the natural log of its number of employ￾ees. We also control for differences between the four high-tech sectors. All four sectors are high-tech indus￾tries but there is still considerable dissimilarity in their R&D and patenting-intensity (OECD, 1992), which can partly be explained by inter-sectoral differences between the nature of knowledge and the regime of appropri￾ability. We included sector dummies to control for these differences. Given the international differences in patent￾ing behavior between companies from the USA, Europe and Japan (OECD, 2001), we control for the national￾ity of the acquiring firm. A National Science Foundation (1998) study shows that the number of patents increased significantly during the period of our study. To control for this we included year dummies. Pre-sample differences in the number of patents of the sample firms should have a positive influence on the innovative performance of the acquiring firm after the M&A. Firms that achieve high levels of innovativeness have, according to Dosi (1988), also a higher chance of maintaining or increasing their level of innovativeness. The firm heterogeneity control variable pre-sample patents is measured as the sum of patents obtained by a firm in the 3 years prior to the firm’s entry into the sample. As already mentioned in the previous section, the rel￾ative size of the acquired knowledge base is calculated by dividing the absolute size of the acquired knowledge base by the size of the acquiring firm’s knowledge base. However, in a few cases, the acquiring firm’s knowl￾edge base is zero for some of the years in the panel data set. In order to calculate the relative size correctly, we follow Hausman et al. (1984) and set these specific val￾ues to one instead of zero and use a dummy variable, called dummy absolute size acquirers, to correct for this solution. 4. Results Appendix A provides the descriptive statistics for all variables.5 There is little correlation between the main 5 We do not report the extensive correlation statistics, this informa￾tion is available from the authors. independent variables, with the expected exception of relatedness of the acquired knowledge base with its squared term. There is also little correlation between the independent variables and the control variables. Only the correlation between pre-sample patents and R&D is rather high (0.784), although lower than reported by Ahuja and Katila (0.89). Since we have to control for unobserved heterogeneity and also because we prefer to follow the Ahuja and Katila (2001) model, we will keep both variables in our model. Table 1 displays the estimation results of the negative binomial model with distributed lag analysis for the post￾M&A innovative performance of the sample firms. We use the full model to discuss our results. The results of the other models are presented in Appendix B. Hypothesis 1 argues that non-technological M&As contribute little to the innovative output of the acquir￾ing firm or that there might even be a negative impact on the post-M&A innovative performance. The indi￾vidual coefficients of the variable for the number of non-technological acquisitions are significant and three out of four show a negative sign. The summed coef- ficient, reflecting the total impact, is negative and significant. Together this provides strong support for Hypothesis 1. We also anticipate that the acquisition of the target firm’s knowledge base will lead to increased economies of scale and scope and a broader knowledge base. Both are expected to have a positive effect on the post-M&A innovative performance (Hypothesis 2). The individual coefficients of the variable absolute size of acquired knowledge base are all significant but show a diverse picture. The absolute size of acquired knowledge base improves post-M&A innovative performance during the first 2 years, but thereafter it has a negative influ￾ence. The summed coefficient is significant and neg￾ative, which also provides no support for Hypothesis 2. Similar to the study of Ahuja and Katila, the total impact of the absolute size of the acquired knowledge base on post-M&A innovative performance is rather small given that a one-unit increase in the absolute size of the knowledge base leads to a 0.02% decrease in the post-M&A innovative performance of the acquiring firm. Hypothesis 3 predicts a negative relationship between the relative size of the acquired knowledge base and the post-M&A innovative performance of the acquir￾ing firm. The individual and summed coefficients are all significant and negative, which clearly supports this hypothesis. Hypothesis 4 states that the acquired knowledge base has to show enough overlap with the acquirer’s already

648M. Cloodt et al. / Research Policy 35 (2006) 642-654Table 1Negative binomial regression with distributed lag analysisVariableFull model3.44 [0.148]***Intercept0.174 [0.012]*No. of non-technological acquisitions (t- 1)0.081 [0.008]*No. of non-technological acquisitions (t- 2)0.069 [0.007]**No. of non-technological acquisitions (t- 3)0.088 [0.012]**No. of non-technological acquisitions (t- 4)0.00025 [0.000002]**Absolute size of acquired knowledge base (t- 1)0.000028[0.000003]"**Absolute size of acquired knowledge base (t- 2)0.000228[0.000008]**Absolute size of acquired knowledge base (t- 3)0.000032[0.000005]**Absolute size of acquired knowledge base (t- 4)0.496 [0.149]**Dummy absolute size acquired knowledge base (t- 1)0.483 [0.145]***Dummy absolute size acquired knowledge base (t 2)0.747 [0.161]***Dummy absolute size acquired knowledge base (t 3)0.727 [0.169]***Dummy absolute size acquired knowledge base (t -- 4)0.253 [0.027]***Relative size of acquired knowledge base (t - 1)0.073 [0.039]Relative size of acquired knowledge base (t 2)0.069 [0.028]*Relative size of acquired knowledge base (t- 3)0.091 [0.031]**Relative size of acquired knowledge base (t- 4)0.797 [0.237]**Relatednessofacquiredknowledgebase (t-1)1.069 [0.197]**Relatedness of acquired knowledge base (t-2)1.084 [0.364]*Relatednessofacquiredknowledgebase(t-3)0.591[0.237]′Relatedness of acquired knowledge base (t-4)0.658 [0.251]**Relatedness of acquired knowledge base sq (t- 1)0.967 [0.226]*Relatedness of acquired knowledge base sq (t-2)0.727 [0.406]Relatedness of acquired knowledge base sq (t-3)Relatedness of acquired knowledge base sq (t- 4)0.117 [0.297]0.106[0.011]***Cultural distance (t -1)0.007 [0.014]Cultural distance (t2)0.022 [0.013]]Cultural distance (t -3)0.115 [0.009]*Cultural distance (t - 4)0.247 [0.031]**Dummy cultural distance (t 1)0.119 [0.038]Dummy cultural distance (t 2)0.163 [0.028]**Dummy cultural distance (t - 3)0.354 [0.027]*Dummy cultural distance (t - 4)0.001546[0.000022]**R&D (t-1)log employees (t-1)0.022 [0.013]0.437 [0.066]***Computers and office machinery0.099 [0.059]Pharmaceuticals0.378 [0.042]***Electronics and communicationsEuropean firms0.054 [0.052]0.293 [0.003]***Presample patents (t 1)N2429log likelihood9217.112Summed coefficients0.273 [0.015]***No. of non-technological acquisitions0.000208 [0.000009]***Absolute size of acquired knowledge base0.486 [0.050]***Relative size of acquired knowledge base3.542 [0.474]**Relatedness of acquired knowledge base2.469 [0.551]***Relatedness of acquired knowledge base sq0.249 [0.015]***Cultural distanceStandard errors are in brackets, year dummies are included but not shown.p<0.10.p<0.05.p<0.01.tp<0.001

648 M. Cloodt et al. / Research Policy 35 (2006) 642–654 Table 1 Negative binomial regression with distributed lag analysis Variable Full model Intercept 3.44 [0.148]*** No. of non-technological acquisitions (t − 1) −0.174 [0.012]*** No. of non-technological acquisitions (t − 2) −0.081 [0.008]*** No. of non-technological acquisitions (t − 3) 0.069 [0.007]*** No. of non-technological acquisitions (t − 4) −0.088 [0.012]*** Absolute size of acquired knowledge base (t − 1) 0.000025 [0.000002]*** Absolute size of acquired knowledge base (t − 2) 0.000028 [0.000003]*** Absolute size of acquired knowledge base (t − 3) −0.000228 [0.000008]*** Absolute size of acquired knowledge base (t − 4) −0.000032 [0.000005]*** Dummy absolute size acquired knowledge base (t − 1) −0.496 [0.149]*** Dummy absolute size acquired knowledge base (t − 2) −0.483 [0.145]*** Dummy absolute size acquired knowledge base (t − 3) −0.747 [0.161]*** Dummy absolute size acquired knowledge base (t − 4) −0.727 [0.169]*** Relative size of acquired knowledge base (t − 1) −0.253 [0.027]*** Relative size of acquired knowledge base (t − 2) −0.073 [0.039]† Relative size of acquired knowledge base (t − 3) −0.069 [0.028]* Relative size of acquired knowledge base (t − 4) −0.091 [0.031]** Relatedness of acquired knowledge base (t − 1) 0.797 [0.237]*** Relatedness of acquired knowledge base (t − 2) 1.069 [0.197]*** Relatedness of acquired knowledge base (t − 3) 1.084 [0.364]** Relatedness of acquired knowledge base (t − 4) 0.591 [0.237]* Relatedness of acquired knowledge base sq (t − 1) −0.658 [0.251]** Relatedness of acquired knowledge base sq (t − 2) −0.967 [0.226]*** Relatedness of acquired knowledge base sq (t − 3) −0.727 [0.406]† Relatedness of acquired knowledge base sq (t − 4) −0.117 [0.297] Cultural distance (t − 1) 0.106 [0.011]*** Cultural distance (t − 2) 0.007 [0.014] Cultural distance (t − 3) 0.022 [0.013]† Cultural distance (t − 4) 0.115 [0.009]*** Dummy cultural distance (t − 1) −0.247 [0.031]*** Dummy cultural distance (t − 2) −0.119 [0.038]** Dummy cultural distance (t − 3) −0.163 [0.028]*** Dummy cultural distance (t − 4) −0.354 [0.027]*** R&D (t − 1) −0.001546 [0.000022]*** log employees (t − 1) 0.022 [0.013]† Computers and office machinery −0.437 [0.066]*** Pharmaceuticals −0.099 [0.059]† Electronics and communications −0.378 [0.042]*** European firms −0.054 [0.052] Presample patents (t − 1) 0.293 [0.003]*** N 2429 log likelihood −9217.112 Summed coefficients No. of non-technological acquisitions −0.273 [0.015]*** Absolute size of acquired knowledge base −0.000208 [0.000009]*** Relative size of acquired knowledge base −0.486 [0.050]*** Relatedness of acquired knowledge base 3.542 [0.474]*** Relatedness of acquired knowledge base sq −2.469 [0.551]*** Cultural distance 0.249 [0.015]*** Standard errors are in brackets, year dummies are included but not shown. † p < 0.10. * p < 0.05. ** p < 0.01. *** p < 0.001

649M. Cloodt et al. / Research Policy 35 (2006) 642-654existing knowledge baseto facilitate the absorption pro-5.Discussion and conclusionscess. Yet, the relatedness ofthe acquired knowledge baseOur studyboth replicates and questions someof theneedsenoughdiversitytomakea substantial contributionto the post-M&A innovative performance.The individ-findings from the single sectorstudyof Ahuja and Katilaualandsummedcoefficientsforthevariablerelatedness(2001), extending into an alternative multi-sectoralof acquiredknowledgebase are significant and positive.context of high-tech industries such as aerospace andThe summed coefficient of the squared term is nega-defense,computers and officemachinery,pharmaceu-tive and significant. The individual coefficients for itsticals,and electronics and communications.Similar tosquaredterm are negativeandsignificant,withtheexcepprevious work,wemade a critical distinction betweention of the last lagged period (t-4). Overall, we findthe effect of non-technological M&As and technolog-support for this hypothesis, indicating that the techno-ical M&As on thepost-M&Ainnovativeperformancelogical relatedness of M&As has a curvilinear impactonof the acquiring firm.Non-technological M&As dothepost-M&Ainnovativeperformanceoftheacquiringnot create additional technological learning or makefirm.any other contribution to the post-M&A innovativeAs far as the control variables are concerned,mostperformance.Apparently,they even have a negativeof our findings are quite straightforward. Size of theinfluence on this performancebecausetheycanhaveacompanies and pre-sample differences in the knowl-disruptive effect on a range of activities andestablishededgebases of the sample firmsboth have a significantorganizational routines of the integrating firm (Hitt etpositiveimpactonpost-M&A innovativeperformanceal.,1996).Managingthe repair of these disruptionsThe region and industry of the acquiring firm haverequiresadditional resources that otherwise could haveanegative influenceonpost-M&A innovativeperfor-been invested in long-term innovative projects.6mance, although the region variable is not significantThe positive impact of technological M&As dependsTheyear dummyfor 1989is not significant and negon afirm'sabilitytointegrate this knowledge and toative.In addition,the year dummies for 1990-1994alter existing routines in the organization of its researchare all significant and negative, relative to the omit-(Capron and Mitchell,2000).Ourstudyrevealssometedyear 1995.This showsthat compared with thelimitations to this potentially positive effect during theprevious years, patenting has significantly increasedactualunification of twofirms.During their integration,in1995.bothfirms havetodeal with a disruption oftheirexistingSomewhat surprisingly, cultural distance has a sigorganizational processes and routines (Gerpott,1995nificant positive impact on thepost-M&A innovativeHaspeslaghandJemison,1991).Themanagementoftheperformance of the acquiring firm, although the individ.company has to devotea large amount ofresources to theual 2-year lagged variableis not significant. This findingintegration of a relativelylargeknowledgebase,leavingsupports some less well-known workthatindicates thatfewerresourcesfortheinnovationprocess itself (Hittetal., 1996). This implies that if the acquired firm's knowl-international M&As can have a positive impact on inno-vativeperformancebecauseitmightforcea companyedgebase is relatively large compared to theacquiringto rethink its innovation strategy in a more internationalfirm, this can have serious consequences for the integra-context (e.g.Hoecklin, 1995). It turns out that highertion of the innovative activities of bothM&A partnersR&Deffortshaveasignificantnegative impacton patent-with a negative impact on the acquirer's innovative per-ing output. Additional analysis with a squared term forformance.this variable, not reported here, does indicate a non-As in the chemical industry,the relatedness of thelinear relationship betweenR&D expenditures and theacquiredknowledgebase is also an importantfactorindependent variable. This finding is consistent with pre-the unification of knowledge bases in a high-tech setviousresearchthatdemonstrates aninverseU-shapedting.Itisadvantageoustotheacquiring firm to obtainrelationshipbetweenR&Dexpendituresandpatentingknowledge in areas that are still somewhat related to its(Hagedoorn and Duysters,2002;Scherer,1984).Forexisting activities(Dosi,1988;Teece,1986).Thisadvan-companies with relatively low levels of R&D expendi-tageisfoundina rathersmoothabsorptionprocessofthetures, an increase in R&D expenditures will result in anincreasein thenumber of theirpatents.However,for6 As already noted,non-technological M&As can stilhaveapositivecompanies that already have a relatively high level ofeffect on the general economic performance of the acquiring firm (seeR&D expenditures, a further increaseof these expen-footnote 3). Our analysis shows that, with all else being equal, the esti-ditures will not lead to a substantial growth in newmated number of patents is relatively greater when a non-technologicalpatents.M&A occurs than when there is no acquisition at all

M. Cloodt et al. / Research Policy 35 (2006) 642–654 649 existing knowledge base to facilitate the absorption pro￾cess. Yet, the relatedness of the acquired knowledge base needs enough diversity to make a substantial contribution to the post-M&A innovative performance. The individ￾ual and summed coefficients for the variable relatedness of acquired knowledge base are significant and positive. The summed coefficient of the squared term is nega￾tive and significant. The individual coefficients for its squared term are negative and significant, with the excep￾tion of the last lagged period (t − 4). Overall, we find support for this hypothesis, indicating that the techno￾logical relatedness of M&As has a curvilinear impact on the post-M&A innovative performance of the acquiring firm. As far as the control variables are concerned, most of our findings are quite straightforward. Size of the companies and pre-sample differences in the knowl￾edge bases of the sample firms both have a significant positive impact on post-M&A innovative performance. The region and industry of the acquiring firm have a negative influence on post-M&A innovative perfor￾mance, although the region variable is not significant. The year dummy for 1989 is not significant and neg￾ative. In addition, the year dummies for 1990–1994 are all significant and negative, relative to the omit￾ted year 1995. This shows that compared with the previous years, patenting has significantly increased in 1995. Somewhat surprisingly, cultural distance has a sig￾nificant positive impact on the post-M&A innovative performance of the acquiring firm, although the individ￾ual 2-year lagged variable is not significant. This finding supports some less well-known work that indicates that international M&As can have a positive impact on inno￾vative performance because it might force a company to rethink its innovation strategy in a more international context (e.g. Hoecklin, 1995). It turns out that higher R&D efforts have a significant negative impact on patent￾ing output. Additional analysis with a squared term for this variable, not reported here, does indicate a non￾linear relationship between R&D expenditures and the dependent variable. This finding is consistent with pre￾vious research that demonstrates an inverse U-shaped relationship between R&D expenditures and patenting (Hagedoorn and Duysters, 2002; Scherer, 1984). For companies with relatively low levels of R&D expendi￾tures, an increase in R&D expenditures will result in an increase in the number of their patents. However, for companies that already have a relatively high level of R&D expenditures, a further increase of these expen￾ditures will not lead to a substantial growth in new patents. 5. Discussion and conclusions Our study both replicates and questions some of the findings from the single sector study of Ahuja and Katila (2001), extending into an alternative multi-sectoral context of high-tech industries such as aerospace and defense, computers and office machinery, pharmaceu￾ticals, and electronics and communications. Similar to previous work, we made a critical distinction between the effect of non-technological M&As and technolog￾ical M&As on the post-M&A innovative performance of the acquiring firm. Non-technological M&As do not create additional technological learning or make any other contribution to the post-M&A innovative performance. Apparently, they even have a negative influence on this performance because they can have a disruptive effect on a range of activities and established organizational routines of the integrating firm (Hitt et al., 1996). Managing the repair of these disruptions requires additional resources that otherwise could have been invested in long-term innovative projects.6 The positive impact of technological M&As depends on a firm’s ability to integrate this knowledge and to alter existing routines in the organization of its research (Capron and Mitchell, 2000). Our study reveals some limitations to this potentially positive effect during the actual unification of two firms. During their integration, both firms have to deal with a disruption of their existing organizational processes and routines (Gerpott, 1995; Haspeslagh and Jemison, 1991). The management of the company has to devote a large amount of resources to the integration of a relatively large knowledge base, leaving fewer resources for the innovation process itself (Hitt et al., 1996). This implies that if the acquired firm’s knowl￾edge base is relatively large compared to the acquiring firm, this can have serious consequences for the integra￾tion of the innovative activities of both M&A partners with a negative impact on the acquirer’s innovative per￾formance. As in the chemical industry, the relatedness of the acquired knowledge base is also an important factor in the unification of knowledge bases in a high-tech set￾ting. It is advantageous to the acquiring firm to obtain knowledge in areas that are still somewhat related to its existing activities (Dosi, 1988; Teece, 1986). This advan￾tage is found in a rather smooth absorption process of the 6 As already noted, non-technological M&As can still have a positive effect on the general economic performance of the acquiring firm (see footnote 3). Our analysis shows that, with all else being equal, the esti￾mated number of patents is relatively greater when a non-technological M&A occurs than when there is no acquisition at all

650M.Cloodt et al. / Research Policy 35 (2006)642-654The fact that industries are dissimilar with respectacquired knowledge base if changes in routines and theorganization of research areincremental (Capronandto their knowledge depreciation rate, explains why theMitchell, 2000; Kogut and Zander, 1992). In addition,analysisofAhuja and Katila(2001)showsapositivecon-the acquisition of related knowledge provides opportu-tribution of the absolute size of the acquired knowledgenitiesfor economies of scale and scope(Gerpott,1995;baseduring alongerpost-M&Aperiod.FirmsoperatingHagedoorn and Duysters,2002:Moweryetal.,1996).in a medium-tech sector, such as the chemicals industry.Both advantages leave the acquiring firm with morehave to deal with less environmental turbulence and aresources forinnovative investments.lowerrateofknowledgedepreciation than firms operat-The above suggests that the acquisition of relateding in one of the high-tech sectors.Therefore, firms inknowledgewillhavethemostpositiveimpactonafirm'sthe chemicals industry can also take a longer post-M&Apost-M&A innovative performance.However,the acqui-time period to evaluate and identify the opportunities forsitionofknowledgethatistoosimilartothe alreadyexist-combining and integrating the acquired knowledge.Foringknowledge base is disadvantageous,as the acquir-companies in the group of high-tech industries that weingfirm will have to bearthe costs of obtaining andstudy,even quite recentknowledge,datingbackacoupletransferring external knowledge without any relevantof years alreadybecomesless valuable,itmayevenbeenrichments of its existing knowledge base (Bartlett,worthless and therefore this knowledge plays a positive1993).Somedegree of differentiation in knowledgerolefor onlya limitedperiod of timeafteranM&Ahastakenplace.7between the acquiring and acquired firm will enrich theacquiring firm's knowledge base, creating opportuni-Finally,the current studyconcentrates on M&As andtheir impact on the acquirer's post-M&A innovative per-tiesfor learning and improved innovativeperformance(Bartlett, 1993; Kogut and Zander, 1992). This sug-formance in several high-tech manufacturing industries.gests that to increase innovative performance throughAs this study bothreplicates and questions some ofM&As,companieshavetotarget firms withmoder-the results of the work by Ahuja and Katila (2001)ately related knowledge bases, avoiding targets withitcouldbe worthwhileto conductfutureresearchinknowledge bases that are either too unrelated or toootherindustries.This research could for instancefocusclosely related. The above mentioned result is clearlyonnon-manufacturing industrieswithalternativemea-relatedto existing research that analyzes therelationshipsures of innovation and technological relatedness thatbetween product-market relatedness and performance,are more appropriate for firms in service industries.where several scholars seem to support the inverted-UIn addition, future research could extend this studyshaped model (e.g.Hoskisson and Hitt,1990; Lubatkinbyexaminingthelongtermpost-M&Aeconomicperand Chatterjee, 1994; Markides, 1992; Palich et al.,formance of the acquiring firm, thereby considering2000:Rumelt.1974.1982)the size of the market and the size of the knowl-In contrast to Ahuja and Katila (2001), our resultsedge base in combination.This would enable an evenmore comprehensive assessment of the total impact ofclearlydemonstratethatinahigh-techsettingtheacqui-M&As.sition of a large absolute knowledge base only con-tributes to improved innovative performance during thefirst couple of post-M&A years.After a few years thiskind of acquisition appears to have a negative influ-7 In addition, this phenomenon can be explained by the length oftimeence. This finding may seem somewhat surprising andit takes for tacit knowledge transfer to occur in comparison to codifieddeserves further attention.knowledge transfer. Immediately after the acquisition it is easier toPrior research suggests that knowledge depreciatestransfer codified knowledge that is transmittable through formal lan-and loses its value over time (Glazer and Weiss,1993:guage such as patents than to transfer tacit knowledge that consists ofthe implicit and non-codifiable accumulation of skills and knowledgeHendersonand Cockburn,1994,1996).Inother words,(Nonaka,1991:Reed and DeFillippi,1990).However,thegradual cre-the knowledge in a given period loses its value in subse-ation ofa single organization facilitates thetransferof tacit knowledgequent periods. The rate at which the value of knowledgeover time (Kogut and Zander, 1992). Firms in the chemical industrydepreciates is likely to vary across industries but it ishave a relatively high degree of codified knowledge and a relativelyespecially highin technology intensiveindustries (Wuytslow degree of tacit knowledge (Cohen et al., 2000).Therefore, firms inthis industry can also take a longerpost-M&A timeperiodto integrateetal.,2004).AccordingtoGlazer and Weiss (1993)intheir crucial codified knowledge.In high-tech industries tacit knowl-industries characterized by high turbulence (i.e. high-edge plays a larger role in the innovation process than it does in thetech industries),the value of knowledge tends to depre-chemical industry (see also Winter, 1987).Hence, in these high-techciate faster because of the high levels of inter-periodindustries the transfer of codified knowledge can play a role for onlyuncertainty.a limited period oftimeafter anM&A has takenplace

650 M. Cloodt et al. / Research Policy 35 (2006) 642–654 acquired knowledge base if changes in routines and the organization of research are incremental (Capron and Mitchell, 2000; Kogut and Zander, 1992). In addition, the acquisition of related knowledge provides opportu￾nities for economies of scale and scope (Gerpott, 1995; Hagedoorn and Duysters, 2002; Mowery et al., 1996). Both advantages leave the acquiring firm with more resources for innovative investments. The above suggests that the acquisition of related knowledge will have the most positive impact on a firm’s post-M&A innovative performance. However, the acqui￾sition of knowledge that is too similar to the already exist￾ing knowledge base is disadvantageous, as the acquir￾ing firm will have to bear the costs of obtaining and transferring external knowledge without any relevant enrichments of its existing knowledge base (Bartlett, 1993). Some degree of differentiation in knowledge between the acquiring and acquired firm will enrich the acquiring firm’s knowledge base, creating opportuni￾ties for learning and improved innovative performance (Bartlett, 1993; Kogut and Zander, 1992). This sug￾gests that to increase innovative performance through M&As, companies have to target firms with moder￾ately related knowledge bases, avoiding targets with knowledge bases that are either too unrelated or too closely related. The above mentioned result is clearly related to existing research that analyzes the relationship between product–market relatedness and performance, where several scholars seem to support the inverted-U shaped model (e.g. Hoskisson and Hitt, 1990; Lubatkin and Chatterjee, 1994; Markides, 1992; Palich et al., 2000; Rumelt, 1974, 1982). In contrast to Ahuja and Katila (2001), our results clearly demonstrate that in a high-tech setting the acqui￾sition of a large absolute knowledge base only con￾tributes to improved innovative performance during the first couple of post-M&A years. After a few years this kind of acquisition appears to have a negative influ￾ence. This finding may seem somewhat surprising and deserves further attention. Prior research suggests that knowledge depreciates and loses its value over time (Glazer and Weiss, 1993; Henderson and Cockburn, 1994, 1996). In other words, the knowledge in a given period loses its value in subse￾quent periods. The rate at which the value of knowledge depreciates is likely to vary across industries but it is especially high in technology intensive industries (Wuyts et al., 2004). According to Glazer and Weiss (1993) in industries characterized by high turbulence (i.e. high￾tech industries), the value of knowledge tends to depre￾ciate faster because of the high levels of inter-period uncertainty. The fact that industries are dissimilar with respect to their knowledge depreciation rate, explains why the analysis of Ahuja and Katila (2001)shows a positive con￾tribution of the absolute size of the acquired knowledge base during a longer post-M&A period. Firms operating in a medium-tech sector, such as the chemicals industry, have to deal with less environmental turbulence and a lower rate of knowledge depreciation than firms operat￾ing in one of the high-tech sectors. Therefore, firms in the chemicals industry can also take a longer post-M&A time period to evaluate and identify the opportunities for combining and integrating the acquired knowledge. For companies in the group of high-tech industries that we study, even quite recent knowledge, dating back a couple of years already becomes less valuable, it may even be worthless and therefore this knowledge plays a positive role for only a limited period of time after an M&A has taken place.7 Finally, the current study concentrates on M&As and their impact on the acquirer’s post-M&A innovative per￾formance in several high-tech manufacturing industries. As this study both replicates and questions some of the results of the work by Ahuja and Katila (2001), it could be worthwhile to conduct future research in other industries. This research could for instance focus on non-manufacturing industries with alternative mea￾sures of innovation and technological relatedness that are more appropriate for firms in service industries. In addition, future research could extend this study by examining the long term post-M&A economic per￾formance of the acquiring firm, thereby considering the size of the market and the size of the knowl￾edge base in combination. This would enable an even more comprehensive assessment of the total impact of M&As. 7 In addition, this phenomenon can be explained by the length of time it takes for tacit knowledge transfer to occur in comparison to codified knowledge transfer. Immediately after the acquisition it is easier to transfer codified knowledge that is transmittable through formal lan￾guage such as patents than to transfer tacit knowledge that consists of the implicit and non-codifiable accumulation of skills and knowledge (Nonaka, 1991; Reed and DeFillippi, 1990). However, the gradual cre￾ation of a single organization facilitates the transfer of tacit knowledge over time (Kogut and Zander, 1992). Firms in the chemical industry have a relatively high degree of codified knowledge and a relatively low degree of tacit knowledge (Cohen et al., 2000). Therefore, firms in this industry can also take a longer post-M&A time period to integrate their crucial codified knowledge. In high-tech industries tacit knowl￾edge plays a larger role in the innovation process than it does in the chemical industry (see also Winter, 1987). Hence, in these high-tech industries the transfer of codified knowledge can play a role for only a limited period of time after an M&A has taken place

651M. Cloodt et al. / Research Policy 35 (2006) 642-654Acknowledgementsgapore for their suggestions and comments on earlierversions of this paper.The authors would like to thank three anonymous ref-erees and participants at the Academy of ManagementAppendixAMeeting.NewOrleans,August,2004and a seminar atthe business school of theNational University of Sin-Table A.1.Table A.1Means and standard deviations for all variablesVariableMeanS.D.(1) Patents46.490139.8800.4001.030(2) No. of non-technological acquisitions (t - 1)0.3500.970(3) No. of non-technological acquisitions (t -2)0.3200.900(4) No. of non-technological acquisitions (t-3)0.3100.880(5) No. of non-technological acquisitions (t -4)443.2002303.190(6)Absolute size ofacquired knowledge base (t-1)371.2102032.950(7) Absolute size of acquired knowledge base (t- 2)338.0901943.130(8)Absolutesizeofacquiredknowledgebase(t-3)(9) Absolute size of acquired knowledge base (t- 4)323.5801887.5600.026(10) Dummy absolute size acquired knowledge base (t- 1)0.1600.0270.160(11)Dummyabsolutesizeacquiredknowledgebase(t-20.0270.160(12) Dummy absolute size acquired knowledge base (t-3)(13) Dummy absolute size acquired knowledge base (t- 4)0.0240.1500.1810.354(14) Relative size of acquired knowledge base (t - 1)(15) Relative size of acquired knowledge base (t- 2)0.1710.3480.1600.339(16) Relative size of acquired knowledge base (t-3)0.1570.336(17) Relative size of acquired knowledge base (t - 4)0.012(18) Relatedness of acquired knowledge base (t- 1)0.0840.0130.086(19) Relatedness of acquired knowledge base (t 2)0.0110.078(20) Relatedness of acquired knowledge base (t-3)0.0110.078(21) Relatedness of acquired knowledge base (t-4)0.0070.076(22) Relatedness of acquired knowledge base (t- 1) sq0.0080.078(23) Relatedness of acquired knowledge base (t-2) sq0.0060.070(24) Relatedness of acquired knowledge base (t - 3) sq.0.0060.070(25) Relatedness of acquired knowledge base (t4) sq.0.2270.687(26)Cultural distance (t-1)(27) Cultural distance (t-2)0.2160.674(28) Cultural distance (t-3)0.2090.6660.1940.648(29)Cultural distance (t-4)0.6200.490(30) Dummy cultural distance (t-1)0.6500.480(31)Dummyculturaldistance (t-2)(32) Dummy cultural distance (t- 3)0.6800.470(33) Dummy cultural distance (t - 4)0.6900.460(34) R&D (t- 1)316.159718.3588.8101.962(35) log employees (t- 1)(36) North America0.7400.4400.1300.340(37) Asia0.1300.340(38) Europe0.0610.240(39) Aerospace and defense0.2200.410(40) Computers and office machinery0.2200.420(41) Pharmaceuticals0.5000.500(42) Electronics and communications124.210368.970(43) Presample patents (t- 1)0.1400.350(44) Year 1989(45) Year 19900.1400.350(46) Year 19910.1400.350(47) Year 19920.1400.350(48) Year 19930.1400.350(49) Year 19940.1400.350(50) Year 19950.1400.350

M. Cloodt et al. / Research Policy 35 (2006) 642–654 651 Acknowledgements The authors would like to thank three anonymous ref￾erees and participants at the Academy of Management Meeting, New Orleans, August, 2004 and a seminar at the business school of the National University of Sin￾gapore for their suggestions and comments on earlier versions of this paper. Appendix A Table A.1. Table A.1 Means and standard deviations for all variables Variable Mean S.D. (1) Patents 46.490 139.880 (2) No. of non-technological acquisitions (t − 1) 0.400 1.030 (3) No. of non-technological acquisitions (t − 2) 0.350 0.970 (4) No. of non-technological acquisitions (t − 3) 0.320 0.900 (5) No. of non-technological acquisitions (t − 4) 0.310 0.880 (6) Absolute size of acquired knowledge base (t − 1) 443.200 2303.190 (7) Absolute size of acquired knowledge base (t − 2) 371.210 2032.950 (8) Absolute size of acquired knowledge base (t − 3) 338.090 1943.130 (9) Absolute size of acquired knowledge base (t − 4) 323.580 1887.560 (10) Dummy absolute size acquired knowledge base (t − 1) 0.026 0.160 (11) Dummy absolute size acquired knowledge base (t − 2) 0.027 0.160 (12) Dummy absolute size acquired knowledge base (t − 3) 0.027 0.160 (13) Dummy absolute size acquired knowledge base (t − 4) 0.024 0.150 (14) Relative size of acquired knowledge base (t − 1) 0.181 0.354 (15) Relative size of acquired knowledge base (t − 2) 0.171 0.348 (16) Relative size of acquired knowledge base (t − 3) 0.160 0.339 (17) Relative size of acquired knowledge base (t − 4) 0.157 0.336 (18) Relatedness of acquired knowledge base (t − 1) 0.012 0.084 (19) Relatedness of acquired knowledge base (t − 2) 0.013 0.086 (20) Relatedness of acquired knowledge base (t − 3) 0.011 0.078 (21) Relatedness of acquired knowledge base (t − 4) 0.011 0.078 (22) Relatedness of acquired knowledge base (t − 1) sq. 0.007 0.076 (23) Relatedness of acquired knowledge base (t − 2) sq. 0.008 0.078 (24) Relatedness of acquired knowledge base (t − 3) sq. 0.006 0.070 (25) Relatedness of acquired knowledge base (t − 4) sq. 0.006 0.070 (26) Cultural distance (t − 1) 0.227 0.687 (27) Cultural distance (t − 2) 0.216 0.674 (28) Cultural distance (t − 3) 0.209 0.666 (29) Cultural distance (t − 4) 0.194 0.648 (30) Dummy cultural distance (t − 1) 0.620 0.490 (31) Dummy cultural distance (t − 2) 0.650 0.480 (32) Dummy cultural distance (t − 3) 0.680 0.470 (33) Dummy cultural distance (t − 4) 0.690 0.460 (34) R&D (t − 1) 316.159 718.358 (35) log employees (t − 1) 8.810 1.962 (36) North America 0.740 0.440 (37) Asia 0.130 0.340 (38) Europe 0.130 0.340 (39) Aerospace and defense 0.061 0.240 (40) Computers and office machinery 0.220 0.410 (41) Pharmaceuticals 0.220 0.420 (42) Electronics and communications 0.500 0.500 (43) Presample patents (t − 1) 124.210 368.970 (44) Year 1989 0.140 0.350 (45) Year 1990 0.140 0.350 (46) Year 1991 0.140 0.350 (47) Year 1992 0.140 0.350 (48) Year 1993 0.140 0.350 (49) Year 1994 0.140 0.350 (50) Year 1995 0.140 0.350