Is it the same as average degree?

No. Average degree counts connections per node across a graph; this measure summarizes neighbors' degrees by the focal nodes' degree classes.

Is it an assortativity coefficient?

No. It provides a degree-dependent profile rather than the single correlation-style summary commonly called degree assortativity.

Can disconnected graphs have this measure?

Yes. Local neighborhoods can be summarized even when the graph contains multiple components.

AI Glossary / Graphs and symbolic AI / Intermediate

Average degree connectivity

Definition

Average degree connectivity describes the average degree of neighbors of nodes that share a particular degree. For a simple unweighted graph, it groups nodes by their number of connections and summarizes how connected their immediate neighbors are within each group.

What it means

Suppose you ask whether well-connected nodes tend to connect to other well-connected nodes. Counting each node's links is insufficient: you also need to examine the nodes at the other ends. Average degree connectivity organizes that information by degree class.

The result is usually a mapping from degree k to a neighbor-degree average, rather than one score for the entire graph or one score for each node. A rising pattern can indicate that higher-degree nodes tend to have higher-degree neighbors; a falling pattern suggests a different mixing pattern.

Despite its name, this is not the minimum number of nodes or edges whose removal disconnects the network. That is a separate family of connectivity measures. Nor does it say that the whole network forms one connected component.

How it works

For a simple undirected, unweighted graph, first count each node's degree. For every node with at least one neighbor, average the degrees of its immediate neighbors. Then group nodes by their own degree and average those local values within each degree class.

Grouping makes it possible to compare structural neighborhoods across low-degree and high-degree nodes. It also hides variation: two nodes in the same class may have very different neighbors even though they contribute to the same summary value. Report class sizes and inspect unusual cases before drawing strong conclusions.

Directed and weighted graphs require explicit choices. Incoming and outgoing connections answer different questions, and weighting can emphasize stronger ties. NetworkX provides parameters for these distinctions, but the exact convention must match the analysis. Isolated nodes have no neighbor average mathematically; software handling of that case should be documented rather than interpreted as an observed neighborhood.

Three practical examples

A supplier-network analyst compares whether firms with many trading relationships tend to connect to similarly diversified firms.
A collaboration study examines whether researchers with few collaborators mainly connect through highly connected colleagues.
An infrastructure team uses the degree-class pattern as one descriptive feature when comparing alternative network layouts, alongside separate resilience checks.

Think of it this way

Think of grouping people by how many acquaintances they have, then asking how many acquaintances their acquaintances have on average. Each group tells a different story about its immediate social surroundings.

The analogy does not make acquaintance counts a measure of friendship quality or personal value. Likewise, a graph statistic reflects the defined edges. Its meaning changes if those edges represent occasional contact, financial transactions or critical technical dependencies.

Worked example

Grouping a five-node path

Consider the undirected path A - B - C - D - E. The endpoint nodes A and E each have degree 1. The middle nodes B, C and D each have degree 2.

A's only neighbor is B, with degree 2, so A's average neighbor degree is 2. E has the same value. Therefore the degree-1 class has average degree connectivity 2.

B's neighbors have degrees 1 and 2, giving 1.5. C's neighbors both have degree 2, giving 2. D's value is 1.5. The degree-2 class therefore has average (1.5 + 2 + 1.5) / 3 = 5/3, approximately 1.67.

The summary is {1: 2, 2: 1.67}. It does not imply that each degree-2 node has the same neighborhood: C differs from B and D. It also does not establish robustness; removing C still separates this path into two parts.

Illustrative case study

Fictional case: Fern Networks compares two internal service maps. An analyst sees that high-degree services tend to connect to other high-degree services and proposes calling the architecture resilient.

A review shows that the statistic describes degree mixing, while many important requests still pass through one shared service. The team retains the connectivity curve as descriptive evidence and adds failure simulations and dependency-path analysis. It also separates monitoring links from operational dependencies because mixing them had inflated some node degrees.

The lesson is that a useful network summary must be tied to the question it actually answers. A pattern of connections can motivate a resilience investigation without proving its outcome.

Common misconceptions and mistakes

It returns a value for every node. Average degree connectivity normally groups results by degree; average neighbor degree provides the node-level view.
A higher value always means a better network. The desirable structure depends on the operational purpose.
The word connectivity proves there are alternate routes. Route redundancy requires other analysis.
A smooth curve guarantees a reliable finding. Small degree classes and incomplete edges can create unstable summaries.
Weighted and unweighted results are directly interchangeable. They can emphasize different relationships.

Business use

This measure helps analysts summarize network structure without listing every node's neighborhood. It can support exploratory work in supply chains, collaboration, infrastructure and recommendation graphs.

Define nodes, edges, observation periods and direction before comparing organizations or time windows. A newly added data source may change the apparent mixing pattern without any underlying operational change. Combine the class averages with sample counts, node-level inspection and measures relevant to the actual business objective. Avoid converting a descriptive structural pattern into an unsupported claim about risk or performance.

Check your understanding

Two degree-3 nodes have average neighbor degrees 2 and 8. What is their degree-class average in the simple unweighted case, and what does it hide?

Answer: Their class average is 5. It hides the substantial difference between the two local neighborhoods, so neither node should automatically be described as having neighbors of average degree 5.

Related terms

Average neighbor degree
Assortativity
Directed graph
Weighted graph
Degree
connected component
network resilience
centrality

Learn next

Draw a small path and a small star. Compute every node's degree, then its average neighbor degree, and finally the degree-class averages. Compare how the two graphs organize low-degree and high-degree nodes. Study degree assortativity next, while keeping it separate from tests of failure tolerance.

Sources

NetworkX documentation, average_degree_connectivity, defines the degree-class statistic and its parameters: https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.assortativity.average_degree_connectivity.html

NetworkX documentation, average_neighbor_degree, defines the underlying node-level neighbourhood measure: https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.assortativity.average_neighbor_degree.html

The worked path and fictional service-map case are original teaching examples.

Previous: Average bitrate Next: Average neighbor degree