Post on 29-Sep-2020
Mining Large-scale Corporate Networks
Frank Takes
LIACS, Leiden UniversityAISSR, University of Amsterdam
Leiden Complex Networks Network — LCN2January 29, 2016
CORPNET — LCN2 1 / 61
Introduction
CORPNET — LCN2 2 / 61
Introduction
CORPNET — LCN2 2 / 61
Introduction
F.W. Takes, Algorithms for Analyzing and Mining Real-WorldGraphs, PhD thesis, Leiden University, 2014.
E.M. Heemskerk and F.W. Takes, The Corporate Elite CommunityStructure of Global Capitalism, in New Political Economy 21(1):90–118, 2016. dx.doi.org/10.1080/13563467.2015.1041483
Leiden Institute of Advanced Computer Science (LIACS)
Amsterdam Institute for Social Science Research (AISSR)
Interdisciplinary research
CORPNET — LCN2 3 / 61
Introduction
F.W. Takes, Algorithms for Analyzing and Mining Real-WorldGraphs, PhD thesis, Leiden University, 2014.
E.M. Heemskerk and F.W. Takes, The Corporate Elite CommunityStructure of Global Capitalism, in New Political Economy 21(1):90–118, 2016. dx.doi.org/10.1080/13563467.2015.1041483
Leiden Institute of Advanced Computer Science (LIACS)
Amsterdam Institute for Social Science Research (AISSR)
Interdisciplinary research
CORPNET — LCN2 3 / 61
Introduction
F.W. Takes, Algorithms for Analyzing and Mining Real-WorldGraphs, PhD thesis, Leiden University, 2014.
E.M. Heemskerk and F.W. Takes, The Corporate Elite CommunityStructure of Global Capitalism, in New Political Economy 21(1):90–118, 2016. dx.doi.org/10.1080/13563467.2015.1041483
Leiden Institute of Advanced Computer Science (LIACS)
Amsterdam Institute for Social Science Research (AISSR)
Interdisciplinary research
CORPNET — LCN2 3 / 61
Introduction
F.W. Takes, Algorithms for Analyzing and Mining Real-WorldGraphs, PhD thesis, Leiden University, 2014.
E.M. Heemskerk and F.W. Takes, The Corporate Elite CommunityStructure of Global Capitalism, in New Political Economy 21(1):90–118, 2016. dx.doi.org/10.1080/13563467.2015.1041483
Leiden Institute of Advanced Computer Science (LIACS)
Amsterdam Institute for Social Science Research (AISSR)
Interdisciplinary research
CORPNET — LCN2 3 / 61
Corporate networks
Networks
Nodes are firms
Edges/links indicate for example:
tradingborrowing/lendingownership(board) interlocks
Aim is to understand:
Corporate controlEconomy at a macro levelCorporate elites
CORPNET — LCN2 4 / 61
Corporate network
US WAL-MART STORES, INC.
US EXXON MOBIL CORP
US GENERAL MOTORS COMPANY
US CVS CAREMARK CORPORATION
CH NESTLE S.A.
GB ROYAL DUTCH SHELL PLC
DE DAIMLER AG
DE SIEMENS AG
US BANK OF AMERICA CORPORATION
US AT&T INC.
US INTERNATIONAL BUSINESS MACHINES CORP
DE VOLKSWAGEN AG
FR TOTAL S.A.
FR GDF SUEZ
FR AXA
DE E.ON GLOBAL COMMODITIES SE
DE E.ON SE
US GENERAL ELECTRIC COMPANY
DE ALLIANZ SE
NO STATOIL ASA
US VERIZON COMMUNICATIONS INC
US VALERO ENERGY CORP
US PROCTER & GAMBLE CO
DE METRO AG
US KROGER CO
GB TESCO PLC
DE BASF SE
US CARDINAL HEALTH INC
US BOEING COMPANY (THE)
DE BAYERISCHE MOTOREN WERKE AG
Figure: Board interlock network of 30 firms.
CORPNET — LCN2 5 / 61
Board interlocks
Board interlock: there is a relationship betweenfirms because they share a board member or director
Vladimir I. Lenin, Imperialism, The Highest Stage ofCapitalism, 1916.
“... a personal union, so to speak, is established betweenthe banks and the biggest industrial and commercialenterprises, the merging of one with another through theacquisition of shares, through the appointment of bankdirectors to the Supervisory Boards (or Boards ofDirectors) of industrial and commercial enterprises, andvice versa.”
CORPNET — LCN2 6 / 61
Board interlocks
Board interlock: there is a relationship betweenfirms because they share a board member or director
Vladimir I. Lenin, Imperialism, The Highest Stage ofCapitalism, 1916.
“... a personal union, so to speak, is established betweenthe banks and the biggest industrial and commercialenterprises, the merging of one with another through theacquisition of shares, through the appointment of bankdirectors to the Supervisory Boards (or Boards ofDirectors) of industrial and commercial enterprises, andvice versa.”
CORPNET — LCN2 6 / 61
Board interlocks
Causes of interlocks:
CollusionCooptation and monitoringLegitimacyCareer advancementSocial cohesion
Consequences of interlocks:
Corporate controlEconomic performanceAccess to resources
M. Mizruchi, What do interlocks do? An analysis, critique, and assessment of research on interlocking directorates, inAnnual review of Sociology 22: 271–298, 1996.
CORPNET — LCN2 7 / 61
Board interlocks
Causes of interlocks:
CollusionCooptation and monitoringLegitimacyCareer advancementSocial cohesion
Consequences of interlocks:
Corporate controlEconomic performanceAccess to resources
M. Mizruchi, What do interlocks do? An analysis, critique, and assessment of research on interlocking directorates, inAnnual review of Sociology 22: 271–298, 1996.
CORPNET — LCN2 7 / 61
GRONTMIJ NV
STICHTING RABOBANK PENSIOENFONDS
ADG DIENSTENGROEP SE
PEER HOLDINGS B.V.
ACHMEA REINSURANCE COMPANY NV
EURETCO HOLDING B.V.
BOSKALIS INTERNATIONAL B.V.
STICHTING NEDERLANDSE PUBLIEKE OMROEP
COOP NEDERLAND U.A.
COOP HOLDING B.V.
ROBECO GROEP NV
FOKKER TECHNOLOGIES GROUP B.V.
ONVZ ZIEKTEKOSTENVERZEKERAAR NV
VEBEGO INTERNATIONAL N.V.
AERCAP HOLDINGS N.V.
Q PARK N.V.
N.V. DELI MAATSCHAPPIJ
NVDU ACQUISITION B.V.
COOPERATIE INTRES U.A.
VANDERLANDE INDUSTRIES HOLDING B.V.
EURETCO B.V.
KONINKLIJKE WESSANEN NV
KONINKLIJKE HASKONING DHV GROEP B.V.
STICHTING YMERE
BCD TRAVEL B.V.
COOP GROOTHANDEL B.V.
VAN LEEUWEN BUIZEN GROEP B.V.
OAD GROEP HOLDING B.V.
ARCHIRODON GROUP N.V.
CALDIC B.V.
SMIT INTERNATIONALE NV
NEDERLANDSE ORGANISATIE VOOR TOEGEPAST-NATUURWETENSCHAPPELIJK ONDERZOEK TNO
SHELL INFORMATION TECHNOLOGY INTERNATIONAL B.V.
VAN WIJNEN HOLDING N.V.
VAN WIJNEN GROEP N.V.
HOOGVLIET SUPER B.V.
MCB NEDERLAND B.V.
BCD TRAVEL SUBHOLDING B.V.
NATIONALE POSTCODE LOTERIJ N.V.
ADG DIENSTENGROEP B.V.
HAVENBEDRIJF ROTTERDAM N.V.
MCB INTERNATIONAL B.V.
DELOITTE HOLDING B.V.
KONINKLIJKE SWETS & ZEITLINGER HOLDING N.V.
SWETS & ZEITLINGER GROUP B.V.
AFG MATERIALS B.V.
GAMMA HOLDING NV
BANK NEDERLANDSE GEMEENTEN NV, BNG
GOUDSE NV (DE)
ONDERLINGE WAARBORGMAATSCHAPPIJ CENTRALE ZORGVERZEKERAARS GROEP AANVULLENDE VERZEKERING ZIEKENFONDS UA
TELEGRAAF MEDIA GROEP N.V.
DELTA LLOYD ZORGVERZEKERING NV
FRIESLAND BANK N.V.
TOMTOM SALES B.V.
ACTION HOLDING B.V.
COOPERATIE AVEBE U.A.
VIT ORGANISATIE B.V.
NATIONALE STICHTING TOT EXPLOITATIE VAN CASINOSPELEN IN NEDERLAND
H.E. INVESTMENTS B.V.
ASML SYSTEMS B.V.
VAN LANSCHOT NV
EWALS HOLDINGS B.V.
CORIO N.V.TENNET TSO B.V.
ZEEMAN GROEP B.V.
AMTRADA HOLDING B.V.
KONINKLIJKE WEGENER N.V.
ACE INNOVATION HOLDING B.V.HUISMAN EQUIPMENT HOLDING B.V.
VOLKSWAGEN PON FINANCIAL SERVICES B.V.
ABN AMRO LEVENSVERZEKERING NV
SPAR HOLDING B.V.
JOH. MOURIK & CO. HOLDING B.V.
RFS HOLLAND HOLDING B.V.
PHILIP MORRIS HOLLAND B.V.
UNIT4 N.V.
AUCTUS HOLDING B.V.
POSTKANTOREN B.V.
VITENS N.V.
HUISMAN EQUIPMENT NETHERLANDS B.V.
EUROPEAN AERONAUTIC DEFENCE AND SPACE COMPANY EADS N.V.
SHELL NEDERLAND B.V.
LYONDELLBASELL INDUSTRIES N.V.
KONINKLIJKE AHOLD NV
KONINKLIJKE PHILIPS N.V.
UNILEVER NV
GASTERRA B.V.
ACHMEA BV
ING VERZEKERINGEN NV
HEINEKEN NV
HEINEKEN HOLDING NV
ESSO NEDERLAND B.V.
AEGON NV
ING GROEP NV
RANDSTAD HOLDING NV SHV HOLDINGS NV
AKZO NOBEL NV
ING BANK NV
RABOBANK NEDERLAND
NIDERA B.V.
KONINKLIJKE KPN NV
COOPERATIE VGZ UA
ZUIVELCOOPERATIE FRIESLANDCAMPINA U.A.
KONINKLIJKE FRIESLANDCAMPINA N.V.
KONINKLIJKE LUCHTVAART MAATSCHAPPIJ N.V.
VION HOLDING N.V.
STICHTING PENSIOENFONDS ABP
KONINKLIJKE DSM N.V.
CRH NEDERLAND B.V.
VION N.V.
CANON EUROPA N.V.
EBN B.V.
ACHMEA ZORGVERZEKERINGEN NV
KONINKLIJKE BAM GROEP NV
TNT EXPRESS N.V.
ABN AMRO GROUP N.V.
GLOBAL MOBILITY HOLDING B.V.
PON HOLDINGS B.V.
COOPERATIE FROMFARMERS U.A.
FORFARMERS B.V.
STMICROELECTRONICS N.V.
COOPERATIEVE INKOOPVERENIGING SUPERUNIE B.A.
BASELL SALES & MARKETING COMPANY B.V.
COOPERATIE MENZIS UA
JUMBO GROEP HOLDING B.V.
CARGILL B.V.
ZILVEREN KRUIS ACHMEA ZORGVERZEKERINGEN NV
NUTRECO N.V.
ESSENT N.V.
ROYAL IMTECH N.V.
VGZ ZORGVERZEKERAAR NV
ENECO HOLDING N.V.
TATA STEEL NEDERLAND B.V.
FRIESLANDCAMPINA INTERNATIONAL HOLDING B.V.
ASML HOLDING N.V.
NEDERLANDSE SPOORWEGEN NV
NS GROEP N.V.
POSTNL N.V.
ASR NEDERLAND NV
AEGON NEDERLAND NV
N.V. NUON ENERGY
TATA STEEL IJMUIDEN B.V.
NATIONALE-NEDERLANDEN LEVENSVERZEKERING MAATSCHAPPIJ NV
CONSTELLIUM N.V.
WOLTERS KLUWER NV
DAF TRUCKS N.V.
EXXONMOBIL CHEMICAL HOLLAND B.V.
CSM NVAGIS ZORGVERZEKERINGEN N.V.
NEDERLANDSCHE BANK NV (DE)
ACHMEA INTERNE DIENSTEN N.V.
KONINKLIJKE BOSKALIS WESTMINSTER NV
BOSKALIS WESTMINSTER DREDGING B.V.
AEGON LEVENSVERZEKERING NV
VAN EERD BEHEER B.V.
FUGRO NV
USG PEOPLE N.V.
RANDSTAD GROEP NEDERLAND B.V.
SBM OFFSHORE N.V.
DELTA LLOYD LEVENSVERZEKERINGEN NV
AEGON ASSET MANAGEMENT HOLDING B.V.
BLOKKER HOLDING B.V.
HOOGWEGT GROEP B.V.
MEDIQ NV.
ACHMEA PENSIOEN-EN LEVENSVEZEKERINGEN NV
ARCADIS NV
SLIGRO FOOD GROUP N.V.
SHELL GLOBAL SOLUTIONS INTERNATIONAL B.V.HEIJMANS NV
GEMALTO N.V.
DRAKA HOLDING N.V.
LYONDELL CHEMIE NEDERLAND B.V.
DELTA N.V.
TBI HOLDINGS B.V.
KONINKLIJKE DE HEUS B.V.DE HEUS ANIMAL NUTRITION B.V.
VODAFONE LIBERTEL B.V.
AALBERTS INDUSTRIES NV
KONINKLIJKE COOPERATIE COSUN U.A.
SNS REAAL NV
NAAMLOZE VENNOOTSCHAP UNIVE ZORG
DETAILRESULT GROEP N.V.
SHELL INTERNATIONAL EXPLORATION AND PRODUCTION B.V.
REMEHA GROUP B.V.
VAN OORD N.V.
ALLIANDER N.V.
PRORAIL B.V.
RAILINFRATRUST B.V.
TENNET HOLDING B.V.
SALYM PETROLEUM DEVELOPMENT N.V.
STORK B.V.
REFRESCO GROUP B.V.
COOPERATIEVE LEVENSMIDDELENGROOTHANDEL NEDERLANDS SPERWERVERBOND U.A.
B.V. SPERWER HOLDING
NEDERLANDSE GASUNIE NV
IZA ZORGVERZEKERAAR NV
ASM INTERNATIONAL NV
COOPERATIE COFORTA U.A.
THE GREENERY B.V.
ENEXIS HOLDING N.V.
NATIONALE-NEDERLANDEN SCHADEVERZEKERING MAATSCHAPPIJ NV
MAMMOET HOLDING B.V.
N.V. LUCHTHAVEN SCHIPHOL
KONINKLIJKE VOPAK N.V.
WAVIN N.V.
BALLAST NEDAM N.V.
DAMEN SHIPYARDS GROUP N.V.
PLUKON FOOD GROUP B.V.
RBS HOLDINGS NV
ENEXIS B.V.REFRESCO HOLDING B.V.
BRUNEL INTERNATIONAL NV
DELTA LLOYD SCHADEVERZEKERING NV
VAN OORD DREDGING AND MARINE CONTRACTORS B.V.
IBM NEDERLAND B.V.
HEMA B.V.
DUTCH LION COOPERATIEF U.A.
DE HEUS NEDERLAND B.V.
DE HEUS VOEDERS B.V.
TKH GROUP N.V.
A-WARE FOOD GROUP B.V.
STICHTING PENSIOENFONDS VOOR DE PREDIKANTEN IN DE PROTESTANTSE KERK IN NEDERLAND
LEASEPLAN CORPORATION NV
TOMTOM NV
ROYAL TEN CATE NV
BROCACEF GROEP N.V.
ANWB B.V.
INTERGAMMA B.V.
TBI BOUW B.V.
MERCK SHARP & DOHME B.V.
DURA VERMEER GROEP N.V.
TUI NEDERLAND HOLDING N.V.
IZZ ZORGVERZEKERAAR NV
ABN AMRO BANK NV
ALLIANZ NEDERLAND GROEP NV
TBI TECHNIEK B.V.
OHRA ZIEKTEKOSTENVERZEKERINGEN NV
MACINTOSH RETAIL GROUP NV
SIEMENS NEDERLAND N.V.
STICHTING PHILIPS PENSIOENFONDS
QIAGEN NV
HOLDING NATIONALE GOEDE DOELEN LOTERIJEN N.V.
SNS BANK N.V.
ONVZ HOLDING BV
STERN GROEP NV
DUTCH FLOWER GROUP B.V.
HERTEL HOLDING B.V.
IHC MERWEDE HOLDING B.V.
Corporate networks
Figure: 400, 000 largest firms globally, plotted based on latitude/longitude.
CORPNET — LCN2 9 / 61
Corporate networks
Figure: Global corporate network: over 1, 000, 000 board interlocks.
CORPNET — LCN2 10 / 61
CORPNET
CORPNET — Corporate Network Governance: Power, Ownershipand Control in Contemporary Global Capitalism
What are the features, origins and power political consequencesof corporate governance networks in modern economic life?
Nature: map and analyze the networkOrigins: uncover generating mechanismsPower: understand how it operates
Work with Eelke Heemskerk and Javier Garcia-Bernardo
CORPNET — LCN2 11 / 61
CORPNET
CORPNET — Corporate Network Governance: Power, Ownershipand Control in Contemporary Global Capitalism
What are the features, origins and power political consequencesof corporate governance networks in modern economic life?
Nature: map and analyze the networkOrigins: uncover generating mechanismsPower: understand how it operates
Work with Eelke Heemskerk and Javier Garcia-Bernardo
CORPNET — LCN2 11 / 61
Corporate network analysis
Apply techniques from (social) network analysis to corporate data
Nodes represent around firms across the globe
Edges denote different relationships:
(Undirected) board interlocks: shared senior level directors(Directed) ownership ties based on shareholder information
Node attributes: country, sector, performance indicators, numberof employees, . . .
Edge attributes: number of interlocks, type of shares, number ofshares, ultimate share percentage, . . .
Data source: ORBIS database
CORPNET — LCN2 12 / 61
Three topics
1 Network topology & centrality large-scale
2 Community detection large-scale
3 Data quality very-large-scale
CORPNET — LCN2 13 / 61
Network topology & centrality
Based on: F.W. Takes and E.M. Heemskerk, Centrality in the Global Network of Corporate Control, forthcoming, 2016.
CORPNET — LCN2 14 / 61
Dataset
ORBIS database (Bureau van Dijk)
Firms listed as “large” or “very large”, and “active”
Personal interlocks at senior management and board level
Snapshot from December 2013
Two-mode network of 971, 891 firms and 3, 272, 523 top executives
579, 924 firms did not have any interlocks
The remaining 391, 967 nodes form the nodes in the one-modeglobal firm-by-firm network
CORPNET — LCN2 15 / 61
Dataset
ORBIS database (Bureau van Dijk)
Firms listed as “large” or “very large”, and “active”
Personal interlocks at senior management and board level
Snapshot from December 2013
Two-mode network of 971, 891 firms and 3, 272, 523 top executives
579, 924 firms did not have any interlocks
The remaining 391, 967 nodes form the nodes in the one-modeglobal firm-by-firm network
CORPNET — LCN2 15 / 61
Topological properties
Global networkNodes 391, 967Edges 1, 711, 968
Density 2.229 · 10−5
Average degree 8.746Clustering coefficient 0.755
Degree assortativity 0.260Components 55, 616
Giant componentNodes 238, 859 nodes (60.9%)Edges 1, 533, 030 (89.5%)
Density 5.374 · 10−5
Average degree 12.83Clustering coefficient 0.751
Degree assortativity 0.202Average distance 7.775
Radius 18Diameter 34
CORPNET — LCN2 16 / 61
Topological properties
Global networkNodes 391, 967Edges 1, 711, 968
Density 2.229 · 10−5
Average degree 8.746Clustering coefficient 0.755
Degree assortativity 0.260Components 55, 616
Giant componentNodes 238, 859 nodes (60.9%)Edges 1, 533, 030 (89.5%)
Density 5.374 · 10−5
Average degree 12.83Clustering coefficient 0.751
Degree assortativity 0.202Average distance 7.775
Radius 18Diameter 34
CORPNET — LCN2 16 / 61
Topological distributions
101
102
103
104
105
0 50 100 150 200 250
fre
que
ncy
degree
degree
Figure: Degree distribution of thegiant component
100
101
102
103
104
105
106
0 10 20 30 40 50 60
fre
qu
en
cy
component size (number of nodes)
component size (number of nodes)
Figure: Component size distribution(excluding giant component)
CORPNET — LCN2 17 / 61
Topological distributions
101
102
103
104
105
106
107
108
109
1010
1011
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
frequency
distance
distance
Figure: Distance distribution
100
101
102
103
104
105
18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
fre
qu
en
cy
eccentricity
eccentricity
Figure: Eccentricity distribution
CORPNET — LCN2 18 / 61
National networksCountry Nodes Density Clust. coeff. Degree assort. Avg. dist. Transnat. factorGB 32,962 0.00067 0.356 0.845 6.63 0.26US 24,802 0.00024 0.228 0.741 6.71 0.48ES 11,102 0.00143 0.156 0.849 6.30 0.25NO 8,963 0.00130 0.173 0.613 5.69 0.40FR 8,896 0.00083 0.170 0.445 6.13 0.77MY 7,878 0.00398 0.115 0.785 4.50 0.07DE 7,224 0.00142 0.320 0.799 8.15 0.63SE 6,656 0.00166 0.430 0.829 6.40 0.79NL 6,083 0.00271 0.225 0.785 7.61 0.84IN 5,911 0.00173 0.047 0.332 4.72 0.20CA 5,439 0.00146 0.072 0.352 5.20 0.52DK 4,517 0.00229 0.163 0.549 5.61 0.78IT 4,483 0.00125 0.198 0.524 7.57 0.88BE 3,264 0.00254 0.123 0.416 5.17 1.57RU 2,939 0.00263 0.102 0.556 6.57 0.08KR 2,802 0.00174 0.124 0.356 5.83 0.05FI 2,626 0.00294 0.174 0.539 5.52 1.11JP 2,605 0.00119 0.113 0.208 7.20 0.21IE 2,497 0.01479 0.178 0.747 5.78 0.39AT 2,142 0.00440 0.273 0.670 5.58 0.79PT 2,120 0.00488 0.138 0.620 5.45 0.56AU 1,897 0.00382 0.085 0.414 4.94 0.58LU 1,484 0.00705 0.196 0.720 6.72 1.55SG 1,472 0.00709 0.080 0.421 4.14 0.90VN 1,393 0.00558 0.090 0.501 4.44 0.01CH 999 0.00620 0.077 0.316 4.78 1.63CN 891 0.00475 0.132 0.465 5.80 1.18KY 642 0.00693 0.098 0.387 5.40 3.90
CORPNET — LCN2 19 / 61
DEUTSCHE PFANDBRIEFBANK AG
COMMERZBANK AG
DEUTSCHE BANK AG
AAREAL BANK AG
IKB DEUTSCHE INDUSTRIEBANK AG
HSBC TRINKAUS & BURKHARDT AG
HYPO REAL ESTATE BANK INTERNATIONAL AGLANDESBANK BERLIN HOLDING AG
MLP AG
WUSTENROT & WURTTEMBERGISCHE
BAYER PHARMA AKTIENGESELLSCHAFT
SIEMENS AG
AXEL SPRINGER SE
VATTENFALL EUROPE AKTIENGESELLSCHAFT
SOLON SE
SALZGITTER AG
MUNCHENER RUCKVERSICHERUNGS-GESELLSCHAFT AKTIENGESELLSCHAFTGENERALI DEUTSCHLAND HOLDING AG
HANNOVER RUECK SE
ALLIANZ SE
AXA KONZERN AKTIENGESELLSCHAFT
VOLKSWAGEN AG
MOBILCOM AKTIENGESELLSCHAFT
PHOENIX AG
AURUBIS AG
JUNGHEINRICH AG
FIELMANN AGMPC MUNCHMEYER PETERSEN CAPITAL AG
CONERGY AG
TAG IMMOBILIEN AG
SYMRISE AG
CONTINENTAL AG
TUI AG
VISCOM AG
DRAGERWERK AG & CO. KGAA
FREENET AG
BASLER AG
H&R AG
NORDEX SE
GLOBAL PVQ SE
JENOPTIK AG
CARL ZEISS MEDITEC AG
GELSENWASSER AG
VOSSLOH AG
BALDA AG
WINCOR NIXDORF AG
DAB BANK AG
COMDIRECT BANK AKTIENGESELLSCHAFT
BHW HOLDING AG
AIXTRON SE
IVG IMMOBILIEN AG
SOLARWORLD AG
RHEINMETALL AG
HENKEL AG & CO. KGAA
E.ON SE
DIS AG
KOLBENSCHMIDT PIERBURG AG
METRO AG
M.A.X. AUTOMATION AG
GERRESHEIMER AG
GEA GROUP AKTIENGESELLSCHAFT
KLOECKNER-WERKE AKTIENGESELLSCHAFT
KLOCKNER & CO SE
ARCANDOR AG
HOCHTIEF AG
MEDION AG
RWE AG
THYSSENKRUPP AG
EVONIK DEGUSSA GMBH
UNITED INTERNET AG
DEUTSCHE LUFTHANSA AG
DEUTZ AG
INDUS HOLDING AG
COLONIA REAL ESTATE AG
BAYER AG
LANXESS AG
ALTANA AG
SOFTWARE AG
MERCK KGAA
T-ONLINE INTERNATIONAL AG
FRAPORT AG
DEUTSCHE BORSE AG
DEUTSCHE WOHNEN AG
BIOTEST AG
SGL CARBON SE
FRESENIUS SE & CO. KGAA
STADA ARZNEIMITTEL AG
CELANESE AG
SAP SE
SYGNIS AG
ENBW ENERGIE BADEN-WURTTEMBERG AG
PRIMION TECHNOLOGY AG
KSB AG
SUDZUCKER AKTIENGESELLSCHAFT MANNHEIM/OCHSENFURT
FUCHS PETROLUB SE
MVV ENERGIE AG
HUGO BOSS AG
ELRINGKLINGER AG
IDS SCHEER AG
PRAKTIKER AG
PORSCHE AUTOMOBIL HOLDING SE
CELESIO AG
DURR AG
MOBILCOM-DEBITEL AG
DAIMLER AG
TAKKT AG
KUKA AG
LOWE AG
WACKER CHEMIE AG
LINDE AG
MAN SE
CONSTANTIN FILM AG
EPCOS AG
WACKER NEUSON SE
MTU AERO ENGINES AG
QIMONDA AG
GIGASET AG
LEONI AG
PUMA SE
ADIDAS AG
GFK SE
PFLEIDERER GMBH
KRONES AG
KOENIG UND BAUER AG
RHON-KLINIKUM AG
MEDIGENE AG
MORPHOSYS AG
PHOENIX SOLAR AG
PROSIEBENSAT.1 MEDIA SE
CONSTANTIN MEDIEN AG
INFINEON TECHNOLOGIES AG
ERGO VERSICHERUNG AG
AXA LEBENSVERSICHERUNG AKTIENGESELLSCHAFT
P039376981
P039376981
P038961035
P038961035
P000920573
P014145074,P161235703P011955373
P014145053,P009012871
P038926198
P011955373
P238497028
P238497028
P014068420
P011421115
P055707185
P116654482
P055707185
P000920573
P014145074
P011750337
P038968505
P223766654
P013909420
P038968505,P009198701P011955379
P039814893
P011955379
P013909420,P014141388
P014141685
P011750337
P011955379,P000857601
P000857601
P001681411,P011955379
P011750337
P038953634
P001681411,P011955379
P038953634
P001681411
P102110608
P177318047
P001074557
P000918132
P058025333
P058025333
P001414649
P011750337
P011750337
P001414649,P000092461
P001906192
P158074524
P014141375,P158074524
P001532121
P014145488
P038952530
P206326058
P158074524,P014139633,P001385312
P104291549,P158074524,P014139633
P206326058
P001408178,P014139633,P010527847
P001532121
P039374384
P158074524,P001385312
P001532121
P011955377,P038952530
P014400189
P107851676
P014139633
P014141492
P014141492
P257302127
P000920573
P014145488
P039376981P001887055
P193907528
P039242894
P039587833
P039635318
P039263990,P048092696
P186136666
P158074524
P158074524
P158074524
P039541129,P041819446,P013146649,P258542133,P044858423,P014141147,P223106707,P121948178,P014141431,P014141160,P014141433,P039581929,P014141437,P014141424,P176117440,P158074524,P038953752
P014141421
P001754437,P014141402
P038968505,P014141408P011955386,P055268765,P014141402
P238723512
P039737960
P014141405,P009436149,P158074524
P158074524
P014141402
P001754437,P071498628
P158074524
P011492479
P055268765P238723512
P011492111
P011955386
P038950285
P038942662
P000817434
P014141402
P039657405
P014144594
P014141402
P000817434
P001754437
P014144594
P064160292
P014141392
P231312619
P000817434
P001532121
P142894582
P001532121
P039463351
P039446397
P014141367
P038954674,P038951390
P013909420
P038938679,P100530927,P038954674,P038951408,P024048928,P039010182,P038951390,P193908195
P014141367
P014141367
P143768861
P038953059
P038954700,P067656975
P014141489
P045851019
P067656975
P014141489
P039570129,P039805939,P014145074,P039805938,P058161279,P039454361
P014145061
P046931328
P038949186
P038949186
P038949186
P011955379
P001059584
P011955373
P011955379
P014141402
P011955379
P055268765
P011955379
P011955386
P014145490
P014145488
P039800046
P001022356
P001022356
P001022356
P038992410
P038952530
P039451197,P039300196,P001372042,P153435435,P262986060,P039554825,P147072272,P262986158,P262986112,P262986111,P262986064,P262986061,P262986118,P262986045,P014139017,P262986046,P262986043,P262986042,P262986197,P262986038,P262986195,P005380953,P001912562,P019055141,P262986213,P039451189,P039451188,P198282831,P147043791,P039251339,P169952647,P262986103,P145695000,P039566602,P262986108,P015548437,P262986201,P262986207,P262986057,P262986054,P262986055,P262986166,P262986050,P262986165,P262986184,P262986031,P013030515
P019200677
P039500550
P001022356
P001022356,P014141372
P019200677
P014141380,P081796726,P014141367,P039377155,P014141370,P014141379,P216686812,P238723512,P000302445,P038949021,P000202880
P158023882
P158023882,P008854667,P014141371,P014141367,P039454466,P216686812,P047166003,P014141370,P038950216
P014141374
P009436271
P000865432
P206326058
P000865432
P009320574,P068688753
P039491604
P068688753
P039479526
P038954674,P038951390
P043617284
P039479481
P043619885
P019200677
P000847826
P142880867
P142880867
P039406760,P132724806,P122051592,P000841605,P129788420,P001194311,P001004604
P039081305
P039975126
P038950451
P039491115
P064045558
P038950451
P068688753
P001017597
P001017597
P064045558
P000718705
P014144594
P014141478
P039555617
P038950451
P001022356
P138443539
P206329708
P014141493
P014141493
P000865432,P038958940,P038923507,P039485366
P039658668
P039043650
P001880133
P019060228
P039268804
P001095070
P039268804
P039565507
P058025333
P039512971
P038940125 P039239355
P039709979
P122285077
P145678748
P001176444
P038953666
P038992387
P001153327,P093250516
P038961035
P142880867
P216679802,P038949231,P038949225,P019259694,P247281636
P206339215
P247281636
P247281636
P247281636
P000722068,P158074524,P014139633
P074578231P011955379
P019038687,P014139633
P064045558,P011955379,P158074524,P001385312
P019038687P000722068
P014139633
P011955379
P061522910
P009318412
P001395491
P014141441
P014144573,P014139633
P158074524
P000722068
P014139633
P158023666
P014290329,P247281636
P014290329,P247281636
P014290329,P247281636
P088624716
P038993902
P038993902
P067656975
P001530309
P147583894
P058155050
P039048002
P038950395
P022372697
P021304691
P022372697
P001414649
P021304691
P001454580
P011750337
P039565508
P204583486
P238497028
P001017597
P000857601,P007442504
P011955379,P157998473
P014141175
P014141175
P007442504
P011955379,P014141129
P055915980
P157998473
P007442504
P039602991
P019038687
P010527996
P068546490
P014139633
P154446383
P038960998
P001532121
P014141493,P014368743
P001213126
P007442504
P014141162
P014141162
P157053662,P041307684
P007442504P014141489
P001945991
P001945991
P064223153
P064223153
P001681411,P011955379 P014141159
P001681411
P157998473
P014141159
P038957448
P014141455P010136509
P022372697
P014141175
P130190463
P163633490
P010519606,P112937077,P009430608,P010519476
P010519606
P038947020
P001154833
P001945991
P070882208
P038949186
P038949186
P021304691
P010858291
P038943750
P014146062
P014146062
P014146062
P039052024
P014350538
P055707185
P038949186
P001555934
P001904263P001904263
P007442504
P001406444
P014141162
P143852726
P259798984
P143818823
P014290329,P247281636
P014290329,P247281636
P010519606
P124245989
P216686812,P014141370,P014141367
P038993902,P014141492,P038953634
P039268804
P014290329,P247281636
P001904263
P046012297
P001681411
P158023882
P064223153
P122033342
P014146062
P014146062
P001554669
P011034822
P014141159
P095942469
P000719218,P000831240,P001860802
P014146062
P014146067
P058699685
P058938047
P039479959
P065699709
P000264537
Work by master student Roberto Lucchese
Findings
Small world phenomenon
Average node-to-node distance
Global network: 7.775National networks: 5.692 (average) or 6.188 (weighted average)
National footprints still visible?
Competing elitesGlobalization
Let’s investigate more complex embeddedness measures!
CORPNET — LCN2 22 / 61
Findings
Small world phenomenon
Average node-to-node distance
Global network: 7.775National networks: 5.692 (average) or 6.188 (weighted average)
National footprints still visible?
Competing elitesGlobalization
Let’s investigate more complex embeddedness measures!
CORPNET — LCN2 22 / 61
Findings
Small world phenomenon
Average node-to-node distance
Global network: 7.775National networks: 5.692 (average) or 6.188 (weighted average)
National footprints still visible?
Competing elitesGlobalization
Let’s investigate more complex embeddedness measures!
CORPNET — LCN2 22 / 61
Centrality
Node centrality: the importance of a node with respect to theother nodes based on the structure of the network
Centrality measure: computes the centrality value of all nodes inthe graph
Degree centrality: number of connectionsCloseness centrality: average distance to all other nodesBetweenness centrality: relative number of times a node is on ashortest path
But what is the ground truth to verify these measures?
Hard to say!Correlate with firm prominence (revenue)?
CORPNET — LCN2 23 / 61
Centrality
Node centrality: the importance of a node with respect to theother nodes based on the structure of the network
Centrality measure: computes the centrality value of all nodes inthe graph
Degree centrality: number of connectionsCloseness centrality: average distance to all other nodesBetweenness centrality: relative number of times a node is on ashortest path
But what is the ground truth to verify these measures?
Hard to say!
Correlate with firm prominence (revenue)?
CORPNET — LCN2 23 / 61
Centrality
Node centrality: the importance of a node with respect to theother nodes based on the structure of the network
Centrality measure: computes the centrality value of all nodes inthe graph
Degree centrality: number of connectionsCloseness centrality: average distance to all other nodesBetweenness centrality: relative number of times a node is on ashortest path
But what is the ground truth to verify these measures?
Hard to say!Correlate with firm prominence (revenue)?
CORPNET — LCN2 23 / 61
Centrality measures compared
Figure: Degree, closeness and betweenness centrality
Source: “Centrality” by Claudio Rocchini, Wikipedia File:Centrality.svg
CORPNET — LCN2 24 / 61
Global vs. National centrality
Global1. US AT&T INC.
2. US 7-ELEVEN INC.
3. GB ROYAL DUTCH SHELL
4. GB ERNST & YOUNG EUROPE
5. KR SAMSUNG ELECTRONICS
6. GB PRICEWATERHOUSECOOPERS
7. CH RAIFFEISEN SCHWEIZ
8. GB KPMG EUROPE
Great Britain1. GB ERNST & YOUNG EUROPE
2. GB PRICEWATERHOUSECOOPERS
3. GB KPMG EUROPE
4. GB ROYAL DUTCH SHELL
5. GB DELOITTE
6. GB JP MORGAN
7. GB EASYJET
8. GB DLA PIPER INTERNATIONAL
CORPNET — LCN2 25 / 61
Global vs. National centrality
Global1. US AT&T INC.
2. US 7-ELEVEN INC.
3. GB ROYAL DUTCH SHELL
4. GB ERNST & YOUNG EUROPE
5. KR SAMSUNG ELECTRONICS
6. GB PRICEWATERHOUSECOOPERS
7. CH RAIFFEISEN SCHWEIZ
8. GB KPMG EUROPE
Great Britain1. GB ERNST & YOUNG EUROPE
2. GB PRICEWATERHOUSECOOPERS
3. GB KPMG EUROPE
4. GB ROYAL DUTCH SHELL
5. GB DELOITTE
6. GB JP MORGAN
7. GB EASYJET
8. GB DLA PIPER INTERNATIONAL
CORPNET — LCN2 25 / 61
Global centrality
Table: Correlation between centrality measures and with firm prominence(revenue), n = 238, 859.
Betweenness Closeness Degree Eigenvector
Betweenness 1.000 0.430 0.521 0.356Closeness 0.430 1.000 0.495 0.902Degree 0.521 0.495 1.000 0.498Eigenvector 0.356 0.902 0.498 1.000
Firm prominence 0.192 0.109 -0.046 0.064
CORPNET — LCN2 26 / 61
National centrality
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
AT AU BE CA
CH
CL
CN
CO
DE
DK
ES FI FR GB
IE IL IN IT JP KR KY LU MX
MY
NL
NO
PT RO
RU
SE SG TH US
ZA
Ra
nk c
orr
ela
tio
n w
ith
firm
pro
min
en
ce
(re
ve
nu
e)
Country
Betweenness centralityCloseness centrality
Degree centralityEigenvector centrality
Figure: Correlation between firm prominence (revenue) and national centrality
CORPNET — LCN2 27 / 61
National vs. global centrality
Centrality persistence: correlation between global centrality (inthe full network) and national centrality (in a partition)
CORPNET — LCN2 28 / 61
Centrality persistence
0
0.2
0.4
0.6
0.8
1
1.2
AT AU BE CA
CH
CL
CN
CO
DE
DK
ES FI FR GB
IE IL IN IT IT JP KR KY LU MX
MY
NL
NO
PT RO
RU
SE SG TH US
ZA
Ce
ntr
alit
y p
ers
iste
nce
Country
Betweenness centralityCloseness centrality
Degree centralityEigenvector centrality
Figure: Centrality persistence for the 35 largest countries.
CORPNET — LCN2 29 / 61
Centrality persistence
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95
No
rma
lize
d lo
g(G
DP
)
Betweenness centrality persistence
AT
AU
BE
CA
CH
CL
CN
CO
DE
DK
ES
FI
FRGB
IE IL
INITIT
JP
KR
KY
LU
MX
MY
NLNO
PTRO
RU
SE
SGTH
US
ZA
best linear fitted model
Figure: Betweenness centrality persistence vs. normalized log(GDP).
CORPNET — LCN2 30 / 61
National vs. global centrality
Partition ranking dominance: the ranking of a partition withinthe full network.Assume S ⊆ V in a graph G = (V ,E ).Assume that a node v ∈ S according to some centrality rankinghas rank r(v) ∈ [0, |V |] in the full ranking of all nodes in V .Partition ranking dominance pcr(S ,V ) is then defined as:
pcr(S ,V ) = 0.5−∑
v∈S r(v)
|S | · |V |
value > 0 means the partition is less central than expectedvalue < 0 means it is more central than expected
CORPNET — LCN2 31 / 61
Partition ranking dominance
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
CH
ZA MX
DE
US
FR SE RU
IT FI LU AT BE SG IN GB
NO
NL
JP CA
ES KR MY
PT IE TH DK
CN
IL CL
RO
AU KY CO
Pa
rtitio
n r
an
kin
g d
om
ina
nce
(b
etw
ee
nn
ess c
en
tra
lity)
Country
Figure: Partition ranking dominance (based on betweenness centrality).
CORPNET — LCN2 32 / 61
Community detection
Based on: E.M. Heemskerk and F.W. Takes, The Corporate Elite Community Structure of Global Capitalism, in NewPolitical Economy 21(1): 90–118, 2016. dx.doi.org/10.1080/13563467.2015.1041483
CORPNET — LCN2 33 / 61
Community detection
Community: set of nodes connected more strongly with eachotherthan with the rest of the network
Community detection algorithms:
Clique-based methodsHierarchical clusteringDivisive algorithms (centrality-based)Modularity maximization algorithms
Country network: aggregate firms from the same country
CORPNET — LCN2 34 / 61
Community detection
Figure: Communities: node subsets connected more strongly with each other
CORPNET — LCN2 35 / 61
Modularity
Modularity: numerical value indicating the quality of a division ofa network into communities
Community: subset of nodes for which the fraction of links insidethe community is higher than expected in a random network
Modularity Q ∈ [0, 1]
Resolution parameter r indicating how “tough” the algorithmshould look for communities
Algorithms optimize the modularity score Q given some r (usinghill climbing, heuristics, genetic algorithms and many moreoptimization techniques)
V.D. Blondel, J-L. Guillaume, R. Lambiotte and E. Lefebvre, Fast unfolding of communities in large networks in Journal ofStatistical Mechanics: Theory and Experiment 10: P10008, 2008.
CORPNET — LCN2 36 / 61
Community detection
CORPNET — LCN2 37 / 61
Community detection
JM
LU
NL
SG
MX
TT
KY
US
PG
AT
GB
ZM
DK
JPMT
HRIT
CZ
ES
FR
AU
BE
EE
SA
NO
ZA
PT TR
CA
MK
GR
CO
SI
PL
BA
TH
UA
CL
IS
HK
KZ
MY
RS
SE
CN
BG
CHLI
ME
VG
CY
IE
TW
LV
RU
FI
DE
BM
VE
RO
IN
IL
SK
HU
NZ
SZ
BN
CDCG
TZ
SV
GH
GAUG
ZWBW
LK
NA
AOPE
MZ
PH
VN
KRLB
PK
NG
LC
PA
RW
MATN
QA
PY
MU
LR
NI
OM
LT
MC
MH
EGKW
YE
LY
SN
JO
SY
TG
KE
ID
IQ
BI
SD
MR
CI
LS
MW
BF
BJ
NEML
GN
UY
HN
DO
AM
CM
MG
NP BT
CF
AE
DZ BH
MO
DM
AD GE
MD
GM
IR
BS
PS
GI
AR
GT
BR
EC
KH
HT
GD
GQ
BB
GY
CV
BO
AL
CR
AI
KG
BZ
FJ
BD
CORPNET — LCN2 38 / 61
Community detection
JM
LU
NL
SG
MX
TT
KY
US
PG
AT
GB
ZM
DK
JPMT
HRIT
CZ
ES
FR
AU
BE
EE
SA
NO
ZA
PT TR
CA
MK
GR
CO
SI
PL
BA
TH
UA
CL
IS
HK
KZ
MY
RS
SE
CN
BG
CHLI
ME
VG
CY
IE
TW
LV
RU
FI
DE
BM
VE
RO
IN
IL
SK
HU
NZ
SZ
BN
CDCG
TZ
SV
GH
GAUG
ZWBW
LK
NA
AOPE
MZ
PH
VN
KRLB
PK
NG
LC
PA
RW
MATN
QA
PY
MU
LR
NI
OM
LT
MC
MH
EGKW
YE
LY
SN
JO
SY
TG
KE
ID
IQ
BI
SD
MR
CI
LS
MW
BF
BJ
NEML
GN
UY
HN
DO
AM
CM
MG
NP BT
CF
AE
DZ BH
MO
DM
AD GE
MD
GM
IR
BS
PS
GI
AR
GT
BR
EC
KH
HT
GD
GQ
BB
GY
CV
BO
AL
CR
AI
KG
BZ
FJ
BD
CORPNET — LCN2 39 / 61
Community detection
JM
LU
NL
SG
MX
TT
KY
US
PG
AT
GB
ZM
DK
JPMT
HRIT
CZ
ES
FR
AU
BE
EE
SA
NO
ZA
PT TR
CA
MK
GR
CO
SI
PL
BA
TH
UA
CL
IS
HK
KZ
MY
RS
SE
CN
BG
CHLI
ME
VG
CY
IE
TW
LV
RU
FI
DE
BM
VE
RO
IN
IL
SK
HU
NZ
SZ
BN
CDCG
TZ
SV
GH
GAUG
ZWBW
LK
NA
AOPE
MZ
PH
VN
KRLB
PK
NG
LC
PA
RW
MATN
QA
PY
MU
LR
NI
OM
LT
MC
MH
EGKW
YE
LY
SN
JO
SY
TG
KE
ID
IQ
BI
SD
MR
CI
LS
MW
BF
BJ
NEML
GN
UY
HN
DO
AM
CM
MG
NP BT
CF
AE
DZ BH
MO
DM
AD GE
MD
GM
IR
BS
PS
GI
AR
GT
BR
EC
KH
HT
GD
GQ
BB
GY
CV
BO
AL
CR
AI
KG
BZ
FJ
BD
CORPNET — LCN2 40 / 61
Community detection
JM
LU
NL
SG
MX
TT
KY
US
PG
AT
GB
ZM
DK
JPMT
HRIT
CZ
ES
FR
AU
BE
EE
SA
NO
ZA
PT TR
CA
MK
GR
CO
SI
PL
BA
TH
UA
CL
IS
HK
KZ
MY
RS
SE
CN
BG
CHLI
ME
VG
CY
IE
TW
LV
RU
FI
DE
BM
VE
RO
IN
IL
SK
HU
NZ
SZ
BN
CDCG
TZ
SV
GH
GAUG
ZWBW
LK
NA
AOPE
MZ
PH
VN
KRLB
PK
NG
LC
PA
RW
MATN
QA
PY
MU
LR
NI
OM
LT
MC
MH
EGKW
YE
LY
SN
JO
SY
TG
KE
ID
IQ
BI
SD
MR
CI
LS
MW
BF
BJ
NEML
GN
UY
HN
DO
AM
CM
MG
NP BT
CF
AE
DZ BH
MO
DM
AD GE
MD
GM
IR
BS
PS
GI
AR
GT
BR
EC
KH
HT
GD
GQ
BB
GY
CV
BO
AL
CR
AI
KG
BZ
FJ
BD
CORPNET — LCN2 41 / 61
Community detection
JM
LU
NL
SG
MX
TT
KY
US
PG
AT
GB
ZM
DK
JPMT
HRIT
CZ
ES
FR
AU
BE
EE
SA
NO
ZA
PT TR
CA
MK
GR
CO
SI
PL
BA
TH
UA
CL
IS
HK
KZ
MY
RS
SE
CN
BG
CHLI
ME
VG
CY
IE
TW
LV
RU
FI
DE
BM
VE
RO
IN
IL
SK
HU
NZ
SZ
BN
CDCG
TZ
SV
GH
GAUG
ZWBW
LK
NA
AOPE
MZ
PH
VN
KRLB
PK
NG
LC
PA
RW
MATN
QA
PY
MU
LR
NI
OM
LT
MC
MH
EGKW
YE
LY
SN
JO
SY
TG
KE
ID
IQ
BI
SD
MR
CI
LS
MW
BF
BJ
NEML
GN
UY
HN
DO
AM
CM
MG
NP BT
CF
AE
DZ BH
MO
DM
AD GE
MD
GM
IR
BS
PS
GI
AR
GT
BR
EC
KH
HT
GD
GQ
BB
GY
CV
BO
AL
CR
AI
KG
BZ
FJ
BD
CORPNET — LCN2 42 / 61
Community detection
JM
LU
NL
SG
MX
TT
KY
US
PG
AT
GB
ZM
DK
JPMT
HRIT
CZ
ES
FR
AU
BE
EE
SA
NO
ZA
PT TR
CA
MK
GR
CO
SI
PL
BA
TH
UA
CL
IS
HK
KZ
MY
RS
SE
CN
BG
CHLI
ME
VG
CY
IE
TW
LV
RU
FI
DE
BM
VE
RO
IN
IL
SK
HU
NZ
SZ
BN
CDCG
TZ
SV
GH
GAUG
ZWBW
LK
NA
AOPE
MZ
PH
VN
KRLB
PK
NG
LC
PA
RW
MATN
QA
PY
MU
LR
NI
OM
LT
MC
MH
EGKW
YE
LY
SN
JO
SY
TG
KE
ID
IQ
BI
SD
MR
CI
LS
MW
BF
BJ
NEML
GN
UY
HN
DO
AM
CM
MG
NP BT
CF
AE
DZ BH
MO
DM
AD GE
MD
GM
IR
BS
PS
GI
AR
GT
BR
EC
KH
HT
GD
GQ
BB
GY
CV
BO
AL
CR
AI
KG
BZ
FJ
BD
CORPNET — LCN2 43 / 61
Community detection
JM
LU
NL
SG
MX
TT
KY
US
PG
AT
GB
ZM
DK
JPMT
HRIT
CZ
ES
FR
AU
BE
EE
SA
NO
ZA
PT TR
CA
MK
GR
CO
SI
PL
BA
TH
UA
CL
IS
HK
KZ
MY
RS
SE
CN
BG
CHLI
ME
VG
CY
IE
TW
LV
RU
FI
DE
BM
VE
RO
IN
IL
SK
HU
NZ
SZ
BN
CDCG
TZ
SV
GH
GAUG
ZWBW
LK
NA
AOPE
MZ
PH
VN
KRLB
PK
NG
LC
PA
RW
MATN
QA
PY
MU
LR
NI
OM
LT
MC
MH
EGKW
YE
LY
SN
JO
SY
TG
KE
ID
IQ
BI
SD
MR
CI
LS
MW
BF
BJ
NEML
GN
UY
HN
DO
AM
CM
MG
NP BT
CF
AE
DZ BH
MO
DM
AD GE
MD
GM
IR
BS
PS
GI
AR
GT
BR
EC
KH
HT
GD
GQ
BB
GY
CV
BO
AL
CR
AI
KG
BZ
FJ
BD
CORPNET — LCN2 44 / 61
Community detection
Figure: Communities: node subsets connected more strongly with each other
Communities in corporate networks have a regional character andfinancial ties are clearly visible
Historical events and cultural similarities between countriescorrelate with interlocks
Outliers and effects of randomization
CORPNET — LCN2 45 / 61
Community detection
Figure: Communities: node subsets connected more strongly with each other
Communities in corporate networks have a regional character andfinancial ties are clearly visible
Historical events and cultural similarities between countriescorrelate with interlocks
Outliers and effects of randomization
CORPNET — LCN2 45 / 61
Computing infrastructure
Server grade hardware
Dual high-frequency CPU architecture with2 × Intel Xeon (Haswell) E5-2643, 6 cores, 12 threads, 3.4GHz
Memory: DDR4-2133 RAM, 24 x 64GB = 1536GB = 1.5TB
Storage: 7TB solid state disk (SDD) storage in RAID6
1GBit uplink to the world
Made possible by the High Performance Computing and Networking (HPCN) fund (summer 2015 call) of the University ofAmsterdam.
CORPNET — LCN2 46 / 61
Data quality
CORPNET — LCN2 47 / 61
Data quality
Previous dataset was from September 2013
CORPNET: study all firms
More than 200 million firms
Are all firms equally important?
Do we have all the firms?
What is the quality of the data?
CORPNET — LCN2 48 / 61
Data quality
Data ModelStatistics
Results
Data quality
Accuracy: the data is trueConsistency: data remains clear and verifiable over timeIntegrity: data has not suffered from corruptionCompleteness: do we have all the data?
We “found” that the Spanish market size was ten times larger thanthe US market: one outlier in the data.
CORPNET — LCN2 49 / 61
Average operating revenue
Figure: Observed average revenue per country for 200 million firms
CORPNET — LCN2 50 / 61
Data quality
Assess firm data quality based on comparing intrinsic factors ofcountries using:
Worldbank data on GDP per capita for each countryEurostat data on the number of firms in each countyDistribution of sum of revenues per country in our data
CORPNET — LCN2 51 / 61
Data quality
INTRINSIC FACTORS
- GDP- Population...
REAL AVERAGE
OBSERVED AVERAGE
GDP per capita
OEC
UMC
NOC
10-3 10010-2 10-1102
106Avera
ge O
pera
ting
Revenue
105
104
103
Figure: Richer countries have larger firms
CORPNET — LCN2 52 / 61
Data quality
DATAQUALITY
OBSERVED AVERAGE
INTRINSIC FACTORS
- GDP- Population...
Data completeness
10-3 10010-2 10-1102
106Avera
ge O
pera
ting
Revenue
105
104
103OEC
UMC
NOC
Figure: Richer countries have better quality
CORPNET — LCN2 53 / 61
Data quality
Rich countries have higher average revenue, but better quality,which decreases the observed average (hard to decouple).
We are interested in the real average (given complete data):
1 Real average ∝ GDPnumber of firms
2 Calculate the effect of intrinsic factors and extrapolate to othercountries
3 Calculate the quality of our global firm data
CORPNET — LCN2 54 / 61
Data quality
The distribution of firm operating revenues follows a lognormaldistribution for 95% of firms, with consistent variance
Larger firms are well-represented. Richer countries have higherdata quality. Higher quality decreases the observed average.
100 101102 103 104 105
Company Turnover (thds dollars)
10-9
10-8
10-7
10-6
10-5
10-4
10-3
10-2
Frequency
100 101102 103 104 105
Company Turnover (thds dollars)
10-9
10-8
10-7
10-6
10-5
10-4
10-3
10-2
100 101102 103 104 105
Company Turnover (thds dollars)
10-9
10-8
10-7
10-6
10-5
10-4
10-3
10-2
100 101102 103 104 105
Company Turnover (thds dollars)
10-9
10-8
10-7
10-6
10-5
10-4
10-3
10-2
Figure: Lognormal distribution and addition of firms
CORPNET — LCN2 55 / 61
Data quality
Figure: Observed average revenue Figure: Estimated average revenue
CORPNET — LCN2 56 / 61
Data quality
Predicted observed
Ob
serv
ed
in d
ata
base
102 103 104 105106102
103
104
105
106
Figure: log(predicted observed) =3.15 log(estimated real) +log(completeness)− 1.05
Figure: Actual completeness of our data
CORPNET — LCN2 57 / 61
Completeness per country
CORPNET — LCN2 58 / 61
Other directions
Revolving doors
Top income compensation
Firms in occupied territories
Public tender, procurement
Relation with patent networks
Exchange Traded Funds (ETFs)
Analysis of particular national networks
CORPNET — LCN2 59 / 61
Conclusion
Big corporate network data provides interesting insight in firmpower and control across the globe
Topological properties, centrality analysis and community detectionreveal regional patterns in the global network
Interpretation of measures is crucial and depends on data quality
We understand the completeness of our 200 million firm dataset,now we can assess the effect on the network
CORPNET has a challenging yet exciting time ahead!Website: http://corpnet.uva.nl
We are open to sharing data, best practices and ideas!
CORPNET — LCN2 60 / 61
Conclusion
Big corporate network data provides interesting insight in firmpower and control across the globe
Topological properties, centrality analysis and community detectionreveal regional patterns in the global network
Interpretation of measures is crucial and depends on data quality
We understand the completeness of our 200 million firm dataset,now we can assess the effect on the network
CORPNET has a challenging yet exciting time ahead!Website: http://corpnet.uva.nl
We are open to sharing data, best practices and ideas!
CORPNET — LCN2 60 / 61
Thank you!
Questions?
http://franktakes.nl
http://corpnet.uva.nl
CORPNET — LCN2 61 / 61