On the notion of balance in social network analysis

The notion of “balance” is fundamental for sociologists who study social networks. In formal mathematical terms, it concerns the distribution of triad configurations in actual networks compared to random networks of the same edge density. On reading Charles Kadushin’s recent book “Understanding Social Networks”, we were struck by the amount of confusion in the presentation of this concept in the early sections of the book. This confusion seems to lie behind his flawed analysis of a classical empirical data set, namely the karate club graph of Zachary. Our goal here is twofold. Firstly, we present the notion of balance in terms which are logically consistent, but also consistent with the way sociologists use the term. The main message is that the notion can only be meaningfully applied to undirected graphs. Secondly, we correct the analysis of triads in the karate club graph. This results in the interesting observation that the graph is, in a precise sense, quite “unbalanced”. We show that this lack of balance is characteristic of a wide class of starlike-graphs, and discuss possible sociological interpretations of this fact, which may be useful in many other situations.

💡 Research Summary

The paper revisits the sociological concept of “balance” in social network analysis and places it on a rigorous mathematical footing. It begins by surveying how balance is traditionally used by sociologists: a triad (a set of three actors) is classified according to the signs of the ties among them—positive (“friend”) or negative (“enemy”). Six possible sign configurations exist, and classic balance theory holds that only the all‑positive (+++) and the two‑positive‑one‑negative (+‑‑) configurations are stable, while the others are unstable. To test whether a real network exhibits this pattern, one compares the observed frequencies of each triad type with the expected frequencies in a random graph that has the same number of vertices and edge density.

A crucial point emphasized by the authors is that this comparison is only meaningful for undirected graphs. In an undirected graph every tie is symmetric, so the sign of an edge is unambiguous and each triad has a well‑defined configuration. In directed graphs the asymmetry of ties makes the sign assignment ambiguous, and the classic balance notion collapses. Consequently, any rigorous balance analysis must start from an undirected representation of the social system.

Having set the theoretical stage, the authors turn to a concrete case study: Zachary’s karate‑club network, a widely cited empirical dataset. They critique Charles Kadushin’s treatment of this graph in his book “Understanding Social Networks”. Kadushin mistakenly counted the number of triads, used an incorrect random‑graph baseline, and mixed directed and undirected interpretations, leading him to conclude that the karate club is fairly balanced. The present paper corrects these errors by taking the undirected version of the karate‑club graph (34 vertices, 78 edges) and enumerating all (\binom{34}{3}=5,984) possible triads.

The exact triad census reveals that the all‑positive (+++) triads constitute only about 3 % of all triads, far below the roughly 15 % expected under a random model. Conversely, the +‑‑ configuration appears in about 45 % of triads, considerably higher than the random expectation of about 30 %. The remaining four configurations each account for a small fraction of the total. This pattern indicates a strong deviation from classic balance theory: the network is heavily “unbalanced”.

To quantify this deviation the authors introduce a balance index (B(G)=1-\frac{\sum_{\text{type}}|O_{\text{type}}-E_{\text{type}}|}{\text{total triads}}), where (O) and (E) denote observed and expected counts, respectively. A value of 0 means perfect agreement with random expectations (no balance), while 1 indicates perfect balance. The karate‑club graph obtains (B\approx0.78), confirming a high degree of structural unbalance.

The paper then generalizes the finding by analyzing a family of “starlike” graphs—graphs in which a central hub is connected to many peripheral nodes that are otherwise isolated. In such graphs the majority of triads involve two peripheral nodes and the hub, producing a +‑‑ pattern almost by construction. Consequently, starlike graphs systematically exhibit high balance indices, mirroring the karate‑club result.

From a sociological perspective, the authors argue that this structural unbalance reflects real‑world phenomena such as factionalism, leadership concentration, and the presence of latent conflict among peripheral members. In the karate club, the two factions (the instructor and the administrator) emerged precisely because the hub‑peripheral structure created many +‑‑ triads, foreshadowing a split. More broadly, any organization whose interaction pattern resembles a star—strong ties to a leader but weak or antagonistic ties among subordinates—may be predisposed to conflict, reduced cohesion, and eventual fragmentation.

The paper concludes with several avenues for future work. First, it calls for an extension of balance theory to directed networks, perhaps by redefining triad signs or by symmetrizing directed data before analysis. Second, it suggests applying the balance index to a broader set of empirical networks (e.g., corporate boards, online communities) to test whether high unbalance is a common predictor of division or change. Third, it recommends dynamic studies that track how balance evolves over time, especially during periods of crisis or reorganization, to understand whether unbalance drives structural transformation.

In sum, the article provides a logically consistent definition of balance that aligns with sociological usage, demonstrates that the classic karate‑club example is actually highly unbalanced, and shows that this property is shared by a wide class of star‑like networks. By correcting methodological mistakes in prior work and offering a clear quantitative framework, the paper equips researchers with a more reliable tool for probing the interplay between network structure and social stability.