Skip to content
Back to search results

Closeness Centrality

Introduced by
Bavelas, A.; formalized by Freeman, L. C.
Year
1950 (formalized 1979)
Type
Node-level centrality index
Family
Network science
Font size

What does it do?

Closeness centrality measures how quickly a node can reach all others in a network by computing the inverse of its average shortest-path distance to every other node.

First described by Bavelas (1950) and formally unified by Freeman (1979), it identifies nodes that can spread information or resources efficiently across the entire graph — not merely nodes with many direct contacts.

The intuition behind the method

Imagine dropping a message at a node and watching how fast it reaches every corner of the network. A node with high closeness sits near the center of the web of connections: it needs fewer hops on average to get to anyone else. A peripheral node, even one with many neighbors, may still be far from large parts of the network. Closeness captures this geography-of-the-network idea: short average distance means high closeness, and high closeness means efficient broadcasting or rapid access to information flowing across the graph.

History

Alex Bavelas introduced the concept of communication centrality based on shortest paths in 1950 while studying group problem-solving efficiency at MIT. Linton Freeman's landmark 1979 Social Networks paper formalized closeness centrality alongside degree and betweenness as one of three canonical node-level centrality measures, providing a unified mathematical framework that became the foundation of modern social network analysis.

When to use it

Use closeness centrality when you want to identify nodes that can most efficiently broadcast information, mobilize resources, or access the rest of the network with minimal intermediaries — for example, locating optimal diffusion seeds, influential communicators, or transit hubs. It is appropriate for connected or near-connected graphs of any size. Do not use it as the primary centrality measure when the network is highly fragmented (many disconnected components), because raw closeness scores become near-zero and meaningless; apply Wasserman-Faust normalization or analyze components separately. Avoid confusing closeness with degree: a node can have many direct ties yet still be peripheral in global reach.

Limitations and considerations

  • Sensitive to graph connectivity: in disconnected networks the raw measure collapses and requires normalization adjustments.
  • Treats all shortest paths as equally important; does not account for the volume of traffic or the probability that a specific path is actually used.
  • Computationally expensive on very large graphs (millions of nodes) because all-pairs shortest-path computation scales poorly without approximation algorithms.
  • Does not distinguish between a node that is close to many weak ties versus one close to a few strong, strategically important ties.

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

Strengths

  • Captures global reachability rather than just local connectivity, making it ideal for studying information flow and diffusion.
  • Normalized score (0 to 1) enables comparison across networks of different sizes.
  • Well-established theoretical foundation with decades of validation in sociology, epidemiology, and computer science.
  • Complements degree and betweenness centrality to give a richer picture of node roles.
  • Efficiently computable with BFS (unweighted) or Dijkstra (weighted) for most empirical networks.

Applications

  • Identifying key information brokers or opinion leaders in organizational and social networks.
  • Selecting optimal seed nodes for maximizing the speed of information or disease diffusion.
  • Ranking transit stations or road intersections by their overall accessibility in transportation networks.
  • Locating influential proteins or genes in biological interaction networks.
  • Detecting influential early adopters in technology diffusion studies.
  • Prioritizing intervention targets in public-health contact networks to slow epidemic spread.

How it works

How it works

1

Represent the network as a graph where nodes are actors (people, organizations, web pages, genes, etc.) and edges encode relationships or interactions. The graph may be undirected or directed; edges may be unweighted (each hop costs 1) or weighted (cost equals edge weight, interpreted as distance or inverse strength).

2

For each node v, compute the shortest-path distance d(u, v) to every other reachable node u using Breadth-First Search for unweighted graphs or Dijkstra's algorithm for weighted graphs. For directed graphs, only paths following edge direction are valid, so in-closeness and out-closeness may differ.

Formula
d(u,v) = \text{shortest path length between nodes } u \text{ and } v
3

Sum all shortest-path distances from v to every other node to obtain v's farness. A small farness means v is close on average to the rest of the network; a large farness indicates peripherality.

Formula
\text{farness}(v) = \sum_{u \neq v} d(u, v)
4

Take the reciprocal of the mean shortest-path distance: C(v) = (n-1) / sum of d(u,v). Multiplying by (n-1) gives a normalized score in (0, 1] so nodes in networks of different sizes can be compared. Higher values indicate a more central, reachable position.

Formula
C(v) = \frac{n-1}{\sum_{u \neq v} d(u, v)}
5

In disconnected graphs a node cannot reach all others, so raw closeness is zero or undefined for unreachable pairs. The Wasserman-Faust normalization extends closeness to disconnected graphs by summing only over reachable nodes and rescaling by the fraction of the graph that is reachable, preventing spurious near-zero scores for isolated components.

Common pitfalls

  • Reporting raw closeness on a network with disconnected components without applying Wasserman-Faust normalization, producing misleading near-zero scores.
  • Treating closeness and degree centrality as interchangeable — high degree does not imply high closeness, and vice versa.
  • Ignoring whether edges are directed: in directed networks, closeness should be computed separately for in- and out-directions.
  • Applying unweighted closeness to a weighted network where edge weights encode meaningful distances, losing information about relationship strength.
  • Comparing raw (unnormalized) closeness scores across networks of different sizes without controlling for n.

Frequently asked questions

How does closeness differ from degree centrality?

Degree centrality counts only direct neighbors, so a node can score high if it has many connections even if those connections are all clustered in one corner of the network. Closeness considers the shortest path to every node in the graph, so it rewards global reachability. A peripheral clique hub may have high degree but low closeness.

What should I do when my network has disconnected components?

Apply the Wasserman-Faust (1994) harmonic or reachability-adjusted normalization, which sums only reachable pairs and rescales by the proportion of the network reachable. Alternatively, analyze the largest connected component separately, or use harmonic centrality (the sum of reciprocal distances), which naturally handles unreachable pairs by assigning them a contribution of zero.

Does closeness work on weighted networks?

Yes. Replace BFS with Dijkstra's algorithm using edge weights as distances (or the inverse of strength/similarity weights if higher weight means closer). The resulting weighted closeness reflects both the topology and the quality of connections.

How do I interpret the normalized closeness score?

A score close to 1 means the node can reach every other node in roughly one hop on average — it sits at the center of the network. A score near 0 means the node is very peripheral. Scores above about 0.6 typically indicate genuinely central positions in most empirical networks.

Is closeness appropriate for directed networks?

Yes, but compute it separately in each direction. In-closeness measures how easily others can reach the focal node (useful for studying node prestige or receptivity), while out-closeness measures how easily the focal node can reach others (useful for studying broadcasting capacity).

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. Bavelas, A. (1950). Communication patterns in task-oriented groups. Journal of the Acoustical Society of America, 22(6), 725–730.

    Open DOI reference
    Copied.