Board Interlock Network Analysis
- Introduced by
- Mark S. Mizruchi; Gerald F. Davis
- Year
- 1996
- Type
- Network-analytic pipeline for inter-firm ties created by shared directors
- Family
- Strategic management / corporate governance and networks
What does it do?
Board interlock network analysis treats the corporate economy as a network in which two firms are tied whenever they share a director, and studies how these interlocking directorates channel information, influence, and the diffusion of practices among companies.
Mizruchi's 1996 Annual Review of Sociology synthesis crystallized the field, distinguishing the determinants of interlocks from their consequences and cataloguing the mechanisms — collusion, cooptation and monitoring, legitimacy, career advancement, and social cohesion — that interlocks have been argued to serve. Davis's 1991 study of how the poison-pill takeover defense spread through the board network gave the perspective its canonical demonstration that corporate practices diffuse along director ties. The method combines two-mode-to-one-mode network construction with positional metrics and diffusion modeling.
The intuition behind the method
A director who sits on several boards is a living conduit between those companies: she carries what she sees in one boardroom into the others. Aggregate millions of such overlaps and the corporate world resembles a social network, with firms as nodes and shared directors as edges. That network matters because it is how ideas, practices, and norms travel between firms that have no formal relationship — when a board considers adopting a new governance device or strategy, the experience of companies its own directors also serve becomes a powerful reference. Board interlock analysis maps this web, measures which firms sit at its center, and tests whether being well-connected, or being exposed to neighbors who already did something, predicts a firm's own choices and outcomes. The poison pill spreading boardroom to boardroom is the textbook case.
History
Research on interlocking directorates dates to early-twentieth-century concerns about financial control and antitrust, and matured through sociological studies of corporate power and bank centrality in the 1970s and 1980s. Mizruchi's 1996 Annual Review of Sociology essay, 'What do interlocks do?', became the field's reference point by systematically separating the determinants of interlocks from their consequences and laying out the competing mechanisms — collusion, cooptation and monitoring, legitimacy, career advancement, and social cohesion — while frankly assessing the thin evidence on outcomes. Davis's 1991 Administrative Science Quarterly study of poison-pill diffusion through the intercorporate network supplied an influential demonstration that practices spread along board ties, anchoring an institutional-diffusion stream. The approach has since expanded with formal network methods, global corporate-control mapping, and renewed attention to gender diversity and elite cohesion.
When to use it
Use board interlock network analysis when your question concerns how firms influence one another, imitate practices, access information, or coordinate, and when you can assemble board membership data across many firms over time. It is well suited to studying the diffusion of governance practices, strategies, or norms; the structure of corporate elites and power; cooptation of resource dependencies onto the board; and how a firm's network position relates to its choices and performance. It is less appropriate when the mechanism of interest is firm-internal rather than relational, when board data are incomplete or the firm set is poorly bounded, or when interlocks are too sparse to form a meaningful network. Because shared directors correlate with shared industry, size, and geography, the design must control for these confounds before reading ties as channels of influence.
Limitations and considerations
- Interlocks correlate with industry, size, and geography, so apparent network effects may reflect confounding rather than influence.
- Mizruchi cautioned that the consequences of interlocks are far harder to establish than their existence, and many proposed functions remain contested.
- Two-mode-to-one-mode projection discards information and inflates clustering, complicating interpretation of the resulting ties.
- Comprehensive, longitudinal board data are demanding to compile, and director turnover makes the network time-sensitive and boundary-dependent.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Recasts the corporate economy as a measurable network, capturing inter-firm relationships invisible to firm-by-firm analysis.
- Provides a rigorous way to test diffusion and imitation of strategies and governance practices, as Davis demonstrated with the poison pill.
- Offers a rich menu of positional metrics (centrality, brokerage) linking network structure to firm behavior and power.
- Bridges sociology and strategy, integrating cooptation, legitimacy, and social-cohesion mechanisms that Mizruchi catalogued.
Applications
- Tracing the diffusion of governance devices such as poison pills, golden parachutes, and board reforms across firms.
- Mapping corporate elite structure and identifying the most central and powerful firms in an economy.
- Studying strategic imitation, where interlocked firms adopt similar diversification, acquisition, or compensation practices.
- Examining cooptation, where firms place resource-controlling partners (banks, suppliers) on their boards to manage dependence.
- Linking a firm's network centrality to access to information, legitimacy, capital, or performance outcomes.
How it works
How it works
The raw material is a two-mode (affiliation) structure recording which directors sit on which boards, encoded as a firm-by-director incidence matrix with a one wherever a person serves on a firm's board. Mizruchi stresses that interlocks are fundamentally relational data, so getting this membership matrix right — defining the set of firms, the time window, and how to treat directors who join or leave — is the foundation. Accuracy here determines everything downstream, since every tie in the firm network is inferred from these shared memberships.
Formula
A\in\{0,1\}^{n\times p},\quad A_{ij}=1\ \text{if director }j\in\text{board of firm }iThe affiliation matrix is projected into a one-mode firm-by-firm network by multiplying it with its transpose, so each off-diagonal entry counts the directors two firms have in common and the diagonal records board size. The result is a weighted, undirected network of interlocks among companies. Mizruchi's review treats this intercorporate network as the object of study, noting that analysts must decide whether to keep tie weights, dichotomize them, or distinguish ties created by inside executives (directional, often interpreted as influence) from those created by neutral outside directors.
Formula
C = A A^{\top},\quad C_{kl}=\#\{\text{directors shared by firms }k,l\},\ \ C_{kk}=\text{board size}With the firm network in hand, the analyst computes positional metrics describing each firm's location and prominence: degree (number of interlock partners), betweenness (bridging between otherwise separate firms), eigenvector or Bonacich centrality (being tied to other well-connected firms), and structural-hole measures of brokerage versus constraint. Mizruchi connects these positions to the proposed functions of interlocks — central firms are better placed for information access, cooptation, and legitimacy. These metrics become the independent variables that translate network structure into testable firm-level predictors.
Formula
c_i^{(eig)}=\tfrac{1}{\lambda}\sum_j C_{ij}\,c_j^{(eig)},\quad \text{degree},\ \text{betweenness},\ \text{constraint}A central use of the network is to model contagion: the spread of a practice as a function of exposure to prior adopters through board ties. Davis's analysis of the poison pill operationalized this by predicting a firm's likelihood of adopting the defense from whether firms it was interlocked with had already adopted, controlling for firm characteristics. This network-exposure or contagion specification — a firm's hazard of adopting rising with the weighted count of adopting neighbors — is the workhorse for showing that corporate practices diffuse socially through directors rather than arising independently.
Formula
\Pr(\text{adopt}_i^{t})=F\!\Big(\alpha+\beta\textstyle\sum_j w_{ij}\,\text{adopt}_j^{t-1}+\boldsymbol{\gamma}'\mathbf{x}_i\Big)Finally, network position and tie exposure are related to firm outcomes and behaviors — adoption of governance devices, strategic imitation, interlock-linked similarity in political contributions, acquisitions, or performance — using regression and event-history models with firm and industry controls. Davis's poison-pill diffusion is the canonical example of board ties predicting behavior, illustrating Mizruchi's point that interlocks can transmit practices and norms regardless of whether each individual director consciously intends to. Interpreting these links, while guarding against confounds like shared industry or geography, is where the analysis pays off substantively.
Formula
Y_i=\beta_0+\beta_1\,\text{centrality}_i+\beta_2\,\text{tie exposure}_i+\boldsymbol{\gamma}'\mathbf{c}_i+\varepsilon_iCommon pitfalls
- Inferring influence or collusion from the mere existence of a tie, ignoring Mizruchi's warning that consequences are hard to demonstrate.
- Failing to control for shared industry, size, and location, so homophily masquerades as network contagion.
- Treating all interlocks as equivalent rather than distinguishing executive-created directional ties from outside-director ties.
- Ignoring the temporal ordering of adoption and tie formation, which is essential for any credible diffusion claim.
Frequently asked questions
How is a board interlock network actually constructed?
You start with a two-mode affiliation matrix recording which directors sit on which boards, then project it to a one-mode firm network by multiplying the matrix by its transpose. Each resulting off-diagonal entry counts the directors two firms share, defining a weighted tie between them. As Mizruchi emphasizes, decisions made here — bounding the firm set, the time window, whether to keep tie weights, and whether to separate directional ties created by executives from neutral outside-director ties — fundamentally shape the network and any conclusions drawn from it.
Do board interlocks actually cause firms to behave alike?
Establishing causation is the field's central challenge. Davis's poison-pill study showed that firms interlocked with prior adopters were more likely to adopt, consistent with diffusion through directors. But Mizruchi cautioned that demonstrating real consequences of interlocks is much harder than documenting that interlocks exist, because shared directors correlate with shared industry, size, and geography. Credible claims require controlling for these confounds and respecting temporal order — neighbors must adopt before the focal firm — so that influence is distinguished from mere homophily.
What do the centrality measures mean for a firm?
Degree centrality counts a firm's direct interlock partners; betweenness captures how often it bridges otherwise disconnected firms; eigenvector or Bonacich centrality rewards being tied to other central firms; and structural-hole measures gauge brokerage versus constraint. Mizruchi links these positions to the proposed functions of interlocks: central, well-connected firms are better placed for information access, cooptation, and legitimacy. In analyses, these metrics serve as predictors testing whether a firm's location in the corporate network shapes its choices, influence, and outcomes.
Related methods
Sources
Mizruchi, M. S. (1996). What do interlocks do? An analysis, critique, and assessment of research on interlocking directorates. Annual Review of Sociology, 22, 271-298.
Open DOI referenceCopied.Davis, G. F. (1991). Agents without principles? The spread of the poison pill through the intercorporate network. Administrative Science Quarterly, 36(4), 583-613.
Open DOI referenceCopied.