Network Centrality n.
The definition
Network Centrality is a family of metrics that rank how structurally important a node is within a network, based purely on the pattern of its connections rather than any personal attribute. Degree, closeness, betweenness and eigenvector centrality each define importance differently, so the same node can rank high on one measure and low on another.
The mechanism
Freeman formalised three foundational conceptions of centrality that still anchor the field: degree, closeness and betweenness. 1 Degree centrality counts a node's direct ties, a raw measure of activity. Closeness centrality measures the average distance from a node to every other node, capturing independence and the efficiency of reach. Betweenness centrality counts how often a node falls on the shortest path connecting other pairs, capturing control over the flow that passes between them.
Bonacich introduced a power-based form of centrality in which a node's standing depends on the standing of the nodes it connects to, not simply on how many ties it holds. 2 A node linked to a handful of highly connected others can outrank a node with many low-status connections, which is why eigenvector centrality often diverges sharply from degree centrality in practice.
Borgatti showed that each centrality measure implicitly assumes a different type of network flow, such as the transfer of a single discrete good versus the replication of information across many ties, so no single measure is universally correct for every network. 3 Morrison and colleagues found that researchers frequently select a centrality metric by convention rather than by matching it to the process actually moving through the network under study. 4
In practice
Two employees within the same organisation can hold identical numbers of contacts yet very different structural power.
Worked example
One employee knows thirty colleagues scattered across low-level, peripheral roles: mailroom staff, junior interns, occasional contractors. Another knows thirty colleagues concentrated among department heads, senior engineers and the people who sit closest to leadership. Degree centrality treats both employees identically, since each holds the same number of ties. Eigenvector centrality ranks the second employee far higher, because their contacts are themselves well-connected and consequently better positioned to pass on resources, information or support.
The count of connections says nothing about their quality; who those connections know is often the more decisive variable.
Why it matters
A meta-analysis spanning 1,699 effect sizes across 147 studies conducted between 2000 and 2022 found that degree, closeness, betweenness and eigenvector centrality were all positively associated with firm performance, though the strength of each relationship shifted over the period studied. 5 Within that broader trend, degree centrality became a weaker predictor of performance over time while eigenvector centrality, being well connected to others who are themselves well connected, became a stronger one. 5
The real-world predictive value of a centrality measure depends on the type of network and the outcome being studied, so applying the wrong measure can misidentify who or what actually holds influence. 5 4 A team that optimises for degree centrality when eigenvector centrality is the better predictor risks investing attention and resources in well-connected but ultimately peripheral people, while overlooking the individuals whose position genuinely concentrates organisational power.
Questions of record
What is the difference between degree, betweenness and eigenvector centrality?
Degree centrality simply counts your direct connections. Betweenness centrality measures how often you sit on the shortest path between other people, giving you brokerage power. Eigenvector centrality weights your connections by how well-connected they are, so knowing a few influential people can outrank knowing many peripheral ones.
Which centrality measure should you use to predict influence or information spread?
Choose the measure that matches how the resource actually moves. Betweenness fits situations where something discrete passes along specific paths, such as a referral. Eigenvector fits situations closer to reputation or information that replicates outward, since it rewards being connected to other well-connected people rather than merely being connected to many.
Does network centrality actually predict performance or influence?
Evidence across many studies links higher centrality to stronger performance outcomes, though the strength of that link varies by measure and has shifted over time. Eigenvector centrality has grown into a more reliable predictor within organisations, while raw connection counts have become progressively less telling on their own.
What is the most common mistake people make when choosing a centrality measure?
The most frequent error is picking a centrality measure out of habit rather than checking what actually flows through the network you are studying. A measure built for one kind of flow, say the spread of a single resource, can badly misrepresent influence in a network where the real process is more like the diffusion of shared information.