An independent study reference written by Dr Phuc V. Nguyen. It is not official subject material — for assessment requirements always follow your subject outline and vUWS.

Networksintermediate

Network density

Q: A 40-person unit and a 400-person unit both average 6 contacts per person on the same relation. What will their densities look like?

The 40-person unit will report roughly ten times the density. Density equals average degree divided by n-1. That is 6 over 39, about 0.154, against 6 over 399, about 0.015, a ratio close to ten. Nothing about behaviour differs. The claim that density cannot be computed without a tie count is wrong as well, because the number of ties follows from average degree and group size.

Q: A research team argues that a dense advice network will make an R&D group more innovative. What is the strongest counter-argument?

Dense networks circulate the same information repeatedly, so the group may be well coordinated and poorly supplied with new ideas. Density buys redundancy. Everyone hears everything, which helps execution and coordination, and it also means most contacts supply information the group already holds. New ideas tend to arrive across sparse connections to different groups, so the innovative structure usually mixes a cohesive core with outward-reaching links.

Density is the share of possible ties that actually exist. Count the ties, divide by the number of pairs, and the result sits between zero and one. A density of one means everyone is connected to everyone. Density is the standard summary of how tightly knit a group is, and it is also the most misused whole-network statistic, because it falls automatically as a group grows. People have limited time, so a 500-person firm cannot sustain the same share of ties as a 10-person team. Comparing densities across networks of different sizes almost always manufactures a false finding.

Why it matters

Density is a fill rate. Imagine a grid with one square for every pair of people, and shade the squares where a tie exists. Density is the shaded fraction. In a small team you can shade most of it. In a large firm the grid grows with the square of headcount while each person still only has time for a few dozen relationships, so the shading thins out no matter how sociable everyone is.

Before you read on — recall

A 40-person unit and a 400-person unit both average 6 contacts per person on the same relation. What will their densities look like?

Formulas

Density of an undirected network

\Delta = \frac{L}{n(n-1)/2} = \frac{2L}{n(n-1)}

Here

L

is the number of observed ties and the denominator is the number of possible pairs. The running six-person team has 6 ties out of 15 possible pairs, so its density is 0.40.

Density of a directed network

\Delta_{\text{dir}} = \frac{L}{n(n-1)}

Direction doubles the number of possible ties, because a tie from

i

j

and one from

j

i

are separate. Using the undirected denominator on directed data overstates density by a factor of two.

Average degree, and how it relates to density

\bar{k} = \frac{2L}{n} = \Delta \, (n-1)

Average degree is how many ties a typical actor holds. It does not shrink automatically with size, so it is the safer of the two when comparing groups. A team of 6 and a team of 60 can both average 2 ties per person while their densities are 0.40 and about 0.034. Safer is not size-free. Time and attention still cap how many ties anyone can hold, so a defensible comparison needs matched boundaries, the same relation definition and a benchmark from a random network of the same size.

Worked examples

Scenario

A consultancy reports that a client head office has density 0.42 while its 300-person operations division has density 0.03, and concludes that head office collaborates about fourteen times more.

Solution

Head office has 18 staff, so 153 possible pairs, and density 0.42 means roughly 64 ties, an average of about 7 contacts each. Operations has 44,850 possible pairs, so density 0.03 means roughly 1,350 ties, an average of about 9 contacts each. On the more size-robust measure operations is slightly better connected, not fourteen times worse. Density fell because the denominator grew, not because behaviour changed. Report average degree, or compare each group only against groups of similar size.

Scenario

Two project teams of exactly 12 people are compared. Both have density 0.30. Management concludes the teams are structurally similar.

Solution

Density is a single average and it is blind to arrangement. Both teams have about 20 ties. In the first, the ties are spread so everybody holds three or four contacts. In the second, six people form a complete clique holding 15 ties among themselves and the other six share five thin links. The first team survives losing anyone. The second falls apart if two people leave. Equal density, opposite resilience. Pair density with a centrality distribution before drawing any conclusion.

Common mistakes

✗Higher density is better. Dense groups share information quickly and they also share the same information. A very dense team tends towards redundancy and is slow to pick up outside ideas. The useful question is whether the density suits the task, with dense structures fitting coordination and sparser structures fitting search.
✗Density can be compared across networks of any size. Density is bounded by human capacity divided by group size, so larger groups score lower almost by construction. Compare average degree instead, or restrict comparisons to groups of similar size, and treat average degree as only more robust to size rather than free of it.
✗Density describes how a group is arranged. It is one ratio and it says nothing about shape. A ring, a star, and a clique surrounded by isolates can all report the same density while behaving completely differently.
✗A low density means the group is not collaborating. It may mean the boundary was drawn too wide, or that the measured relation is rare by nature. A network of "co-authored a report with" will always be sparser than one of "has spoken to", and the two densities are not comparable.

Revision bullets

•Density is observed ties divided by possible ties, bounded between 0 and 1
•Undirected denominator is n(n-1)/2; directed is n(n-1)
•Density falls as group size rises, so raw cross-size comparison misleads
•Average degree, 2L divided by n, is the more size-robust alternative, not a size-free one
•Running example: 6 ties out of 15 pairs gives 0.40
•Equal density can hide completely different structures

Quick check

A 40-person unit and a 400-person unit both average 6 contacts per person on the same relation. What will their densities look like?

A research team argues that a dense advice network will make an R&D group more innovative. What is the strongest counter-argument?

Connected topics

Fitness for Use Actors and Ties Degree Centrality Closeness Bridges and Cliques Network Study

More in Networks

What SNA Is Weighted Ties Betweenness

Sources

Wasserman & Faust (1994), Ch. 4
Wasserman, S., & Faust, K. Social Network Analysis: Methods and Applications. Cambridge University Press, 1994.
Defines density for directed and undirected relations and discusses its dependence on network size.
Newman (2018)
Newman, M. E. J. Networks. 2nd ed. Oxford University Press, 2018.
Treats density and mean degree together and shows why mean degree is the stable quantity as networks grow.

How to cite this page

Dr. Phil's Quant Lab. (2026). Network density. Derivatives Atlas. https://phucnguyenvan.com/concept/ba-network-density