Presentation is loading. Please wait.

Presentation is loading. Please wait.

Porous and Fuzzy Boundaries: A Network Approach to Corporate Diversification* David Knoke University of Minnesota Universiteit van Tilburg June 28, 2007.

Similar presentations

Presentation on theme: "Porous and Fuzzy Boundaries: A Network Approach to Corporate Diversification* David Knoke University of Minnesota Universiteit van Tilburg June 28, 2007."— Presentation transcript:

1 Porous and Fuzzy Boundaries: A Network Approach to Corporate Diversification* David Knoke University of Minnesota Universiteit van Tilburg June 28, 2007 *Based on a research paper co-authored with Emanuela Todeva and Donka Keskinova. Sabbatical support provided by the University of Minnesota College of Liberal Arts.

2 Blurred Boundaries Researchers periodically not the “imprecision of industry definition and the ‘fuzziness’ of industry boundaries in economic environments characterized by product differentiation and technological change” (Venkatraman and Thomas 1988:546). “Industry and market boundaries are porous and ‘fuzzy’ especially where globalization is taking place” (McGee, Thomas and Pruett 1995:261). A colorful example is Vivendi SA –active in music, video games, television, film, publishing, telecoms, and Internet – whose current incarnation involved 2000 merger of Seagram, Canal+, and Vivendi; a spin-off of original core water and waste companies; and sale of Universal Studios to NBC in 2006.

3 Corporate Diversification Corporate diversification theories in finance economics and strategic management examine origins, trends, and financial consequences of diversified firms. An “age-old question”: does diversification – business units in different industries controlled by a single firm – create or destroy shareholder value compared to focused firms? (Martin & Sayrak 2003:38) A diverse-focused dichotomy or count of the number of SIC industries obscures complex structural relations linking firms and industries, and fails to investigate whether particular combinations of industries differentially affect firm behaviors and performance outcomes. Firms embedded within specific industrial network configurations may experience competitive advantages or disadvantages relative to firms located in alternative structural arrangements.

4 A Network Approach We test two research hypotheses: H1: The affiliation network reveals two discrete types of firm clusters, (1)Diversified-industry firms operating in two or more industries (2)Focused-industry firms concentrating on a single industry H2: Diversified-industry clusters explain additional variation in firm financial performance above the additive effects of conventional industrial classifications. To better understand blurred boundaries arising from corporate diversification in the Global Information Sector (GIS), we apply social network concepts and methods to reveal the structural relations among its industries and firms & to explain their effects.

5 Affiliation Networks An affiliation network can be displayed either as a bipartite graph, or as a gxh affiliation matrix ( A ) whose i,j entry indicators whether actor i participated in event j. Its hxg transpose matrix ( A’ ) shows whether event j attracted actor i. Formally, a pair of elementary sets connected by a (0-1 binary or ordinal) relation: Set N of g nodes (“actors”): N = {n 1, n 2,..… n g } Set M of h nodes (“events”): M = {m 1, m 2, … m h } L nondirected lines join the gxh ordered pairs of nodes An affiliation network consists of two-mode data, different sets connected by relations between but not within each set. If the two sets are “actors” and “events,” elements within each mode are indirectly tied, via common links to the other mode. Familiar examples of affiliation networks include: persons belonging to voluntary associations; social movement activists participating in protest events; firms creating strategic alliances; nations signing trade and military treaties.

6 Duality of Persons & Groups Ronald Breiger’s (1974) classic article on the duality of persons and groups discussed: (1) actor-actor connections occurring through their co-membership or co-attendance at the same events; and (2) event- event connections via the overlap or interlocks with shared actors. These two dual networks can be created by either pre- or post-multiplying an affiliation network and its transpose to create two one-mode matrices: AA’ is a gxg symmetrical matrix; its main diagonal entries show the number events in which an actor is affiliated; its off-diagonal elements are the number of events in which a row & column pair jointly participated. A’A is an hxh symmetrical matrix whose main diagonal entries show the number actors participating in the row event; its off-diagonal elements are the number of actors affiliated with a particular pair events. Both dual matrices may be analyzed as one-mode networks, measuring such properties as size, density, reachability, and cohesion. Interpretations of co- memberships must recognize that entities are indirectly connected, and that the specific identities of those indirect paths cannot be known from the dual matrix (e.g., we know the number of events a pair attended but not which events).

7 The Global Information Sector The GIS is based on the North American Industrial Classification System information sector (51) of firms producing & distributing info commodities 511 Publishing Industries (except Internet) 512 Motion Picture and Sound Recording Industries 515 Broadcasting (except Internet) 517 Telecommunications 518 Internet Service Providers, Web Search Portals, Data Proc Services 519 Other Information Services + 334 Computer and Electronic Product Manufacturing Using 2005 Fortune and Forbes lists, we found 275 corporations active in at least one of the 33 five-digit GIS industries (median = 2.00, mean = 2.55). NAICS codes from Thomson and Datamonitor. Firm revenues ranged from Pixar Studio’s $300 million to Nippon Telegraph & Telephone’s $101 billion.

8 NAICS Subsectors and Industries in the Global Information Sector ___________________________________________________________________________________________ CodeIndustry NameAbbreviationN ___________________________________________________________________________________________ 334 Computer and Electronic Product Manufacturing 33411Computer and Peripheral EquipmentComputer51 33421Telephone ApparatusTeleApp23 33422RadioTelevision Broadcasting and WirelessBroadcast20 33429Other Communications EquipmentCommunic24 33431Audio and Video EquipmentAV21 33441Semiconductor and Other Electronic ComponentsSemicond73 33451Navigational Measur, Electromedical & Control Inst.Navigat16 33461Manufacturing Reproduc Magnetic & Optical MediaReprod14 511 Publishing 51111Newspaper Publishers News17 51112Periodical PublishersPeriod23 51113Book PublishersBook16 51114Directory and Mailing List PublishersDirectory14 51119Other Publishers OthPub1 51121Software PublishersSoftware45 512 Motion Picture and Sound Recording 51211Motion Picture and Video ProductionMovie13 51212Motion Picture and Video DistributionMovieDist4 51213Motion Picture and Video ExhibitionMovieExh2 51219Postproduction Services and Other IndustriesPostProd3 51222Integrated Record Production DistributionRecord1 51223Music Publishers Music7 515 Broadcasting 51511Radio Broadcasting Radio7 51512Television BroadcastingTV27 51521Cable and Other Subscription Programming Cable21 517 Telecommunications 51711Wired Telecommunications CarriersWired55 51721Wireless Telecommun Carriers except SatelliteWireless58 51731Telecommunications Resellers TCResell25 51741Satellite Telecommunications Satellite27 51751Cable and Other Program Distribution CableDist6 51791Other Telecommunications OtherTC36 518 Internet Service Providers, Web Search Portals, and Data Processing Service 51811Internet Service Providers and Web Search PortalsISP19 51821Data Processing Hosting and Related ServicesDataProc29 519 Other Information Services 51911News SyndicatesSyndic3 51912Libraries and Archives Library1 ___________________________________________________________________________________________

9 Measuring Similarity Semiconductors 1 0 1 Wireless Telecoms 0 a3a3 b 55 c 70 d 148 Newspapers 1 0 1 TV 0 a 17 b 17 c7c7 d 242 In the two-mode 275 x 33 firms-by-industries binary matrix, a cell entry of 1 indicates a row firm operates in the column industry, and 0 indicates absence. For all pairs of columns we computed a 33 x 33 matrix of Jaccard similarity coefficients, the ratio between the size of an intersection to the size of a union for two industries. The higher a Jaccard value, the greater the overlap among the firms in a pair of GIS industries: Jaccard = (a / (a + b + c)) Jaccard = (17 / (17 + 17 + 7)) = 0.41Jaccard = (3 / (3 + 55 + 70)) = 0.02

10 Clustering Industries The next two figures display a hierarchical cluster analysis of the 33 GIS industry similarities (complete-link criterion) and a multidimensional scaling plot (stress = 0.24) with contiguity lines around the six diversified-industry clusters and three singletons. Shown in the following two figures are cluster and MDS analyses of the dual 275 x 275 firm-by-firm matrix of Jaccard coefficients. 1.Bottom clusters are mostly equipment manufacturing (NAICS industries in subsector 334) and telecommunication industries (517), whose proximity implies stronger ties among these industries than to other parts of the Global Information Sector. 2.Presence of software industry (511) inside the cluster with computer manufacturing, navigational equipment, and reproducing media, and the presence of data processing (518) among the telecoms reveal some heterogeneity within those two diversified- industry clusters. 3.The three clusters at the top also exhibit substantial industry heterogeneity, which remains even if the large cluster of industries in the publishing, motion picture, and broadcasting subsectors were divided into two subclusters (dotted line).

11 Dendogram from Hierarchical Cluster Analysis of 33 Industries (Ordinal Scale)

12 Multidimensional Scaling of Jaccard Coefficients among 33 GIS Industries

13 Clustering Firms Next figure is a cluster analysis of 275 x 275 firm-by-firm matrix of Jaccards. 1.The 15 focused-firm clusters, labeled in boldface capitals, each have only a single dominant industry, with no other industry prevalent among least half its member firms. 2.The 11 diversified-firm clusters, labeled in hyphenated lower case letters, have between two and five additional industries in which half or more of their member firms participate. The MDS plots intercluster proximities, calculated as weighted path lengths. (Cell counts are normalized within each matrix row to add to 1.00, then multiplied by the matrix transpose, producing a 24 x 24 cluster-by-cluster matrix. Higher values indicate greater similarity of a pair of firm clusters’ ties to all 33 industries.) 1.Four of 5 groups of firm clusters include focused and diversified industries. 2.Two groups at the upper left involve mixtures of publishing and mass media clusters, respectively. 3.Large heterogeneous group on the right side combines four focused- industry with four diversified-industry clusters of firms. 4.Also in the large group are both clusters of telecom apparatus- communication equipment manufacturers, separated from the adjacent group containing the telecom service-provider clusters.

14 Summary of Hierarchical Cluster Analysis of 275 GIS Firms __________________________________________________________________________________________________________________ Firm Clusters’ Main IndustriesNSome Prominent Firms __________________________________________________________________________________________________________________ 1. MOVIEEX2Regal Entertainment 2. AV3Maxtor, Philips 3. BROADCAST8Agilent, Matsushita, Qualcomm 4. news-tv8Daily Mail, Dow-Jones, Gannett, NY Times, Singapore Press 5. teleapp-commun -semicond7Alcatel, Cisco, Ericsson, Lucent, Nortel 6. TV7DirecTV, Fuji TV, Tokyo Broadcasting, Tribune 7. PERIOD3Primedia, VNU 8. WIRELESS14Comcast, EchoStar, Portugal Telecom, Sprint-Nextel, Telus 9. semiconductor-teleapp-communic10Intel, Nokia, Motorola, Sanyo, Siemens, Sumitomo 10. SEMICONDUCTOR39Kyocera, Mitsubishi, Taiwan Semiconductor, Texas Instruments 11. satellite-wireless-wired- othertc-tcresell-dataproc17Bell Canada, CBS, France Telecom, KDDI, NTT, Telecom Italia 12. wired-othertc17China Unicom, Reuters, Telecom Indonesia, Telenor, Vodafone 13. movie-tv10Disney, News Corporation, Time Warner, Viacom, Vivendi 14. cable-tv12BSkyB, Liberty Global, ITV, Washington Post 15. COMPUTER19Acer, Benq, Bull, Dell, Hewlett-Packard, Hitachi, SanDisk 16. computer-semiconductor12Canon, LSI Logic, Nvidia, Oki, Samsung, Toshiba 17. book-period11Axel Springer, McGraw-Hill, Pearson, Reader’s Digest 18. satellite-tcresell-wired-wireless6AT&T, BellSouth, Hellenic Telecom, Qwest, Telstra, Verizon 19. tcresell-wireless-wired21Alltel, Carso Global, China Netcom, Pakistan Telecom, Turkcell 20. DATAPROCESS12Atos, EDS, First Data, NCR, Unisys, Xerox 21. software-computer-reprod14Apple, Fujitsu, Microsoft, Oracle, SAP, Seagate, Sony, Sun 22. SOFTWARE17Adobe, Autodesk, Avaya, CA, Infosys, Intuit, Siebel, VeriSign 23. ISP5Belgacom, Google, Yahoo 24. DIRECTORY3Dex, Dun & Bradstreet 25. NAVIGATIONAL7Lexmark, Ricoh, Scientific-Atlanta 26. OTHPUB1Seat-Pagine __________________________________________________________________________________________________________________

15 Multidimensional Scaling of Weighted Path Distances among 24 Firm Clusters

16 Explaining Firm Performance Next two tables show ANCOVAs for 17 firm performance indicators, controlling for age, # employees, 32 NAICS industry dummies, and 12 diversified-industry firm clusters from the preceding cluster analysis. 1.Ten industrially diversified firm clusters have significantly effects in one or more equations. Relative to the focused firms, some diversified firms performed better (e.g., total assets, dividend per share), while others performed worse (e.g., net income, ROI). 2.Bottom panel reports F-ratios for tests of differences in R 2 s compared to equations without the 12 diversified-industry firm clusters. All show increased R 2 of 1.5 - 8.3%. In seven instances they boosted the additive R 2 by 20-59%. Thus diversified-industry clusters account for additional variation in firm financial performance beyond that attributable to additive effects of the NAICS industry classification. Numerous opportunities to extend structural analysis: to other economic sectors, with longitudinal data, additional firm outcome measures, etc. North American Product Classification System may soon allow three-mode networks of products-by-firms-by-industries. Then we can test not only whether boundaries are porous & fuzzy, but whether they’re also squishy!



Download ppt "Porous and Fuzzy Boundaries: A Network Approach to Corporate Diversification* David Knoke University of Minnesota Universiteit van Tilburg June 28, 2007."

Similar presentations

Ads by Google