Introduction

For this research I wanted to analyze how well connected the corporate boards of American corporations are and in which ways. To accomplish this goal I have gathered the complete board memberships of nine large companies: Netflix, Amazon, Walmart, Exxon, Berkshire Hathaway, Apple, Nvidia, Coca-Cola, and JPMorgan Chase. This list of individuals was expanded by finding and documenting all of the other corporate board memberships those people also hold concurrently and have held in the past. The goal of this past-inclusive data set is to better track track the influence of corporate culture and capture the ongoing relations which are not completely severed when a director changes boards.

This data was collected using https://www.marketscreener.com/.

Network Information

Order = 112

Size = 859

Undirected Network

No other Attributes encoded

This analysis aims to identify the most highly connected people and companies with this network. These are actors with a strong ability to influence others, build a cohesive culture, and maintain connections throughout the business community.

The importance of this can be appreciated in the scale of financial power which these people either control directly or have some guiding influence over. Further, its importance, can be realized by looking at how many people are directly effected by their decisions, which would be the workers at this firms and their customers. This sample of companies includes many industries which ensures that those of us affected are numbered in the hundreds of millions.

Background

Interlocking directorates are a classic mechanism for corporate coordination and diffusion of norms. We expect from the literature on this topic that network analysis is the appropriate tool for determining the influence and information brokerage opportunities of nodes in this data set.

Network theory provides some tests:

High degree indicates many direct ties (broad immediate reach).

High eigenvector centrality suggests being connected to other well-connected people (indirect influence).

High betweenness indicates brokerage opportunities connecting otherwise distant groups.

Coreness (k-core) identifies the most mutually embedded subnetwork the “core” of dense mutual connections.

Methodology

Data preparation and cleaning -

Merged current and past board membership CSVs, removed empty rows/columns, and produced a cleaned edgelist (person, company). Built an undirected bipartite igraph and assigned vertex types (people vs companies) with bipartite_mapping().

Conversion to one‑mode projection -

Built the people overlap matrix as People × People, set diagonal to zero, and used graph_from_adjacency_matrix(..., weighted=TRUE) to produce the people projection. Removed isolates, after projection).

Measures -

Centralities: degree, eigenvector, hub/authority, betweenness, closeness, ego-reach at k=2 and k=3.

Core-periphery: computed coreness (k-core) and extracted the maximum k-core subgraph.

Community detection: used Girvan–Newman (edge betweenness clustering) and Louvain for comparison.

Bicomponents: decomposed into bicomponents to inspect dense subgraphs.

Structural holes: computed Burt’s constraint (and inverse constraint as brokerage potential).

Visualization: computed using the igraph package.

Findings

K-Core’s in the Network

A closer look at the largest K-Core

This densest core of the network reveals something interesting. It is mostly populated by the board members of JPMorgan Chase, which relates some information about how the banking and investment banking sectors are integrated into the larger economic framework. Facilitating and benefiting from the economic acticity of many other diverse industries.

Nodes within this group with notable ties to other companies besides JPMorgan include: Ana O’shea, a Spanish Banking Leader, Maria Lagomasino, who sits on the boards of Coca-Cola and the Walt Disney corporation, and Ginni Rometty who has deep ties with IBM.

Centrality Scores for the whole graph

Top 10 unique across centrality measures
Name Degree Eig Hub Authority Closeness Reach_2 Reach_3 Betweenness
Maria Lagomasino 35 1.000 1.000 1.000 0.529 0.820 1 332.784
Steve Burke 30 0.891 0.891 0.891 0.541 0.892 1 407.073
Brad Smith 30 0.727 0.727 0.727 0.524 0.829 1 281.154
Bela Bajaria 28 0.620 0.620 0.620 0.512 0.793 1 357.968
Leslie Kilgore 28 0.314 0.314 0.314 0.529 0.856 1 386.928
Ana Patricia O’Shea 27 0.897 0.897 0.897 0.512 0.811 1 123.918
Ann Mather 27 0.386 0.386 0.386 0.509 0.829 1 275.599
Timothy Flynn 27 0.609 0.609 0.609 0.496 0.739 1 232.540
Todd Combs 27 0.886 0.886 0.886 0.487 0.721 1 162.788
Carolyn Everson 26 0.636 0.636 0.636 0.512 0.838 1 256.076

These top ten individuals by combined centrality measures occupy clearly central and influential positions in the people‑projection of the board network. They have the highest direct connectivity (degree) and also score strongly on eigenvector/hub/authority measures, which indicates not only many ties but ties to other well‑connected actors. Closeness values are uniformly very small, and every top individual has Reach 3 of 1, showing that each can reach the entire network within three steps; this implies rapid potential for information or norm diffusion originating from these actors.

Betweenness centrality is more heterogeneous across the top ten: some central actors serve as important brokers linking otherwise distant subgroups, while others are highly embedded in the core but less essential as bridges. Practically, this means influence is concentrated among a small set of tightly interconnected directors, yet different actors play distinct roles, some as broad connectors and others as key gatekeepers or bridges.

Maria Lagomasino (Coca-Cola, Disney), Steve Burke (Berkshire, JPMorgan) are people with a large capacity to both diffuse information throughout the network with high connected peers and have a large reach to the broader network. They combine both influence and brokerage power.

The character of the network as implied by the Reach 3 statistic is that this well knit community will quickly and effecticly be able to spread information and cultural beliefs. Possibly, informing the able of those who lead major companies to be able to act with some form of implicit uniformity.

Girvan-Neuman (Edge Betweenness)

The Investment and Amazon clusters sit centrally and are densely connected to several other clusters, suggesting that directors linked to these sectors act as major hubs tying multiple industries together. By contrast, the Nvidia and Oil & Gas clusters appear more peripheral and internally cohesive, many ties within the cluster but fewer ties outward, implying greater internal cohesion and fewer cross‑industry board ties for those groups.

Banking also sits at a crossroads between other major industries, further supporting the proposition that finance serves enabling and liaison roles within the network.

##  Group 1: n=24
##    Warren Buffett, Kenneth Chenault, Wally Weitz, Steve Burke, Charlotte Guyman, Sue Decker, Chris Davis, Tom Murphy, Meryl Witmer, Gregory Abel
##    ...
##  Group 2: n=14
##    Bela Bajaria, Reed Hastings, Jay Hoag, Gregory Peters, Richard Burton, Anne Sweeney, Strive Masiyiwa, Susan Rice, Elinor Mertz, Bradford Smith
##    ...
##  Group 3: n=10
##    Jeff Bezos, Jamie Gorlick, Indra Nooyi, Patricia Stonesifer, Wendell Weeks, Jonathon Rubinstein, Edith Cooper, Daniel Huttenlocher, Andrew Ng, Keith Alexander
##  Group 4: n=11
##    Brad Smith, Alex Gorsky, Jamie Dimon, Mark Weinburger, Ginni Rometty, Phebe Novakovic, Daniel Pinto, Linda Bammann, Mellody Hobson, Michele Buck
##    ...
##  Group 5: n=13
##    Greg Penner, Carla Harris, Thomas Horton, Timothy Flynn, Cesar Conde, Randall Stephenson, Marissa Mayer, Sarah Friar, Doug McMillon, John Furner
##    ...
##  Group 6: n=13
##    Darren Woods, Michael Angelakis, Joseph Hooley, Greg Garland, Steven Kandarian, Maria Dreyfus, Alexander Karsner, John Harris, Kaisa Hietala, Larry Kellner
##    ...
##  Group 7: n=15
##    James Gorman, Darica Rice, David Darroch, Michael Froman, Amy Chang, Mary Barra, Calvin McDonald, Robert Iger, Arthur Levison, Susan Wagner
##    ...
##  Group 8: n=12
##    Melissa Lora, Mark Stevens, Tench Coxe, Aarti Shah, Dawn Hudson, Harvey Jones, Robert Burgess, Stephen Neal, Brooke Seawell, Persis Drell
##    ...

Bicomponents

Bicomponents for this network cannot be found, the network’s high density has eliminated articulation points on which the network would be divided. Although this does not give us any insight into sub-communities or brokerage opportunities between otherwise unconnected groups it does tell us that the network is high interconnected. With many possible and redundant avenues for information to pass along.

This test was run on the two-mode version of the network to prevent the expansion of ties that occurs when the people-to-people network is projected from the two mode original data set.

Burts Constraint

Burt’s constraint is roughly the extent to which a node’s contacts are redundant (0 = no redundancy, 1 = fully constrained). Low constraint = non‑redundant contacts and a greater ability to broker information between otherwise disconnected nodes in the network.

We see here in the Burts Constraint table that some of the least constrained nodes have access to many other unique nodes, with scores < 1. However, looking deeper into the table it can be seen that even those that score the lowest (around .8) are still relatively unconstrained. Although, those people lower on the constraint list may also be contained by poor access on the other measures. This tells use that the network is high dense and well connected. Which provides many, if not most, of the people with the opportunity to both share and receive information from a broad range of unique sources.

Conclusions

This analysis of the interlocking directorate network built from nine focal companies and their past/current shared board memberships reveals a highly connected, small‑world structure with a clear core of influential actors and a set of identifiable sectoral subcommunities.

Core and reach: A relatively small core of directors combines high degree and eigenvector centrality and can reach the entire network within three steps.

This indicates rapid potential for diffusion of information, norms, and governance practices from core actors to the broader board population.

Brokerage: Several individuals have low Burt constraint and high brokerage potential, indicating they bridge otherwise less‑connected parts of the network. These actors (at firms such as Coca-Cola, Disney, and JP Morgan) are structurally positioned to mediate or gate information across clusters.

Community structure: Girvan–Newman reveal clusters that generally map onto industry or firm groups (Investment and Amazon clusters are central and highly connected; Nvidia and Oil & Gas clusters are more internally cohesive and relatively peripheral). However, the clusters are not isolated from each other.

This research attempts to retest some of the findings of “Finance Capital and Capitalist Class Integration in the 1990s: Networks of Interlocking Directorships in Canada and Australia” by Carroll & Alexander (1999). But fails to capture the scope of the original research. Further research could reveal if co-occurrence on a corporate board effects the likelihood of serving together in the future. Limited sample size and dataset fineness prevent that from being investigated here.

This project does support the general conclusions of Carroll & Alexander. That the tight knit community that makes up those who serve on the boards of the largest corporations make up a cohesive bloc. With strong internal ties to one another, and a myriad of paths by which to communicate and share information and cultural affinity.