Skip to content
Back to search results

Centrality Analysis — Degree, Betweenness, Eigenvector

Introduced by
Linton C. Freeman
Year
1979
Type
Descriptive / exploratory network measure family
Font size

What does it do?

Centrality analysis is a family of network-analytic measures, formalized by Freeman (1979), that quantifies the structural importance of individual nodes within a graph.

Each centrality index captures a distinct mechanism of influence: degree centrality reflects direct connectivity, betweenness centrality identifies nodes that broker information flow, closeness centrality captures proximity to all others, and eigenvector centrality (along with PageRank) rewards connection to highly connected neighbors.

The intuition behind the method

Not every node in a network is equally important. A city that many roads pass through is more critical to traffic than one that sits at the end of a dead-end street, even if both cities have the same number of direct road links. Centrality measures translate this intuition into numbers. Degree centrality simply counts how many neighbors a node has. Betweenness centrality counts how often a node sits on the shortest path between every other pair, revealing gatekeepers. Closeness centrality measures how quickly a node can reach everyone else. Eigenvector centrality weights each neighbor's contribution by how central that neighbor is, so being connected to influential nodes matters more than sheer numbers. Each measure answers a different question about who or what is most important in the network.

History

Freeman (1979) provided the field's foundational conceptual clarification, unifying three previously scattered ideas — degree, betweenness, and closeness — under a single theoretical framework for social network centrality. Borgatti (2005) extended the framework by linking each centrality type to specific models of network flow, giving the measures a clearer theoretical grounding. Eigenvector centrality had earlier roots in the spectral analysis of matrices and was later operationalized for networks; Google's PageRank (1998) is its best-known applied descendant.

When to use it

Centrality analysis applies whenever relational data can be represented as a graph and the question concerns which nodes are structurally most important. It is used in social network analysis, bibliometrics, biology, transportation, and information systems research. No normality assumption is required, and the method is suitable for cross-sectional and longitudinal network data. The network should be connected, or components should be analysed separately, so that path-based measures such as betweenness and closeness are well-defined. At least 10 nodes are needed for scores to be interpretable, and at least 20 for centrality rankings to be statistically meaningful. Weighted networks require a theoretically grounded weight definition before centrality is computed.

Limitations and considerations

  • Betweenness centrality has computational complexity O(VE), making it expensive on very large networks.
  • Path-based measures (betweenness, closeness) are undefined or require special treatment in disconnected graphs.
  • Centrality scores are sensitive to the choice of edge-weight definition in weighted networks; a poorly chosen weight distorts every index.
  • With fewer than 20 nodes the measures lack statistical power and should be treated as exploratory descriptions rather than inferential results.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Each index captures a distinct mechanism of structural importance, giving a multidimensional portrait of the network.
  • No distributional assumptions are required; the method works on any graph regardless of variable type.
  • Degree and eigenvector centrality scale efficiently to very large graphs; degree centrality is computed in O(V+E).
  • Widely supported across disciplines — from sociology and biology to computer science — making results broadly communicable.

Applications

  • Identifying key opinion leaders and information brokers in social and organizational networks.
  • Mapping scientific influence by computing centrality on co-authorship or citation networks.
  • Locating critical infrastructure nodes in transportation, power-grid, and internet topology networks.
  • Prioritizing vaccination or monitoring targets in epidemiological networks by pinpointing high-betweenness spreaders.
  • Detecting gatekeepers in supply-chain and financial networks to assess systemic risk.

How it works

How it works

1

A graph is assembled from the relational data: nodes represent entities (people, institutions, web pages) and edges represent the relationships or interactions between them. The analyst decides whether edges are directed or undirected and whether they carry weights. In directed networks, in-degree and out-degree are computed separately and interpreted differently. The validity of centrality scores depends entirely on whether the graph structure is a faithful representation of the real-world relationships under study.

2

The analyst chooses which centrality index — or combination of indices — matches the substantive question. Degree centrality is appropriate when raw connectivity is of interest. Betweenness centrality is chosen when the focus is on brokerage and information gatekeeping, at the cost of higher computational effort (O(VE)) on large graphs. Closeness centrality is meaningful when the network is connected or when components are analysed separately. Eigenvector centrality or PageRank is selected when prestige and recursive influence matter. For weighted networks the weight definition must be theoretically meaningful before any index is computed.

3

Each selected index is calculated over the graph. Degree centrality is normalized by dividing raw degree by (n-1). Betweenness centrality counts shortest paths through each node and is normalized by the total number of node pairs. Closeness centrality is the reciprocal of the average shortest-path distance from a node to all reachable nodes. Eigenvector centrality is solved as the principal eigenvector of the adjacency matrix and requires iteration to convergence. Nodes are ranked by each score, and the distribution of scores is inspected for concentration.

4

The resulting centrality scores and rankings are interpreted in the substantive context of the study. High-degree nodes are well-connected hubs; high-betweenness nodes are structural bridges or brokers; nodes with high closeness can spread information quickly; high-eigenvector nodes are embedded in influential neighborhoods. Multiple indices are compared to reveal nodes that rank highly on one dimension but not another, which is often the most theoretically interesting finding. Networks with fewer than 20 nodes yield centrality scores that are statistically thin and should be treated as purely descriptive.

Common pitfalls

  • Applying betweenness or closeness to a disconnected graph without first decomposing it into components yields undefined or misleading scores for isolated nodes.
  • Treating all five centrality indices as interchangeable — each answers a different question, and selecting the wrong index for the substantive hypothesis is a common error.
  • Ignoring edge direction in directed networks collapses in-degree and out-degree into a single number, losing information about the asymmetry of relationships.
  • Interpreting centrality scores in networks with fewer than 20 nodes as if they were statistically robust — below this threshold the scores are descriptive only and should not be used to draw inferential conclusions.

Frequently asked questions

Which centrality measure should I use?

The choice depends on your substantive question. Use degree centrality to identify nodes with the most direct connections. Use betweenness centrality to find brokers who control information flow between parts of the network. Use closeness centrality to identify nodes that can reach all others quickly. Use eigenvector centrality or PageRank when the prestige of a node's neighbors matters as much as their number. Reporting two or three complementary indices and comparing the rankings is usually more informative than relying on a single measure.

Can centrality analysis be applied to directed networks?

Yes. In a directed network, in-degree and out-degree are computed and interpreted separately: in-degree reflects how often a node is chosen or cited by others, while out-degree reflects how many others it reaches or references. Betweenness and closeness also have directed variants. Eigenvector centrality on directed graphs typically uses the dominant left or right eigenvector depending on whether prestige flows to or from a node.

What is the minimum network size needed for meaningful results?

The statwise registry specifies a minimum of 10 nodes for any centrality computation, and meaningful statistical interpretation requires at least 20. Below 20 nodes the ranking of a handful of nodes does not generalise, and descriptive network statistics are the more honest choice.

How do I handle a disconnected graph?

Path-based measures — betweenness and closeness — are undefined between nodes in different components. The standard approach is to analyse each connected component separately, or to restrict closeness to reachable nodes and normalize accordingly. Degree and eigenvector centrality do not require connectivity and can be computed on the full graph.

Related methods

Sources

  1. Freeman, L.C. (1979). Centrality in Social Networks: Conceptual Clarification. Social Networks, 1(3), 215-239.

    Open DOI reference
    Copied.
  2. Borgatti, S.P. (2005). Centrality and Network Flow. Social Networks, 27(1), 55-71.

    Open DOI reference
    Copied.