Sociology and Social Networks

Prell’s Social Network Analysis

Prell’s book is the most sociologically informed of the three, though a fundamental, and largely achieved, goal of its author is to make it accessible to a wider audience. The book targets the newcomer and is suitable for use in introductory undergraduate courses. Starting with a well-documented outline of the development of the field, the book grounds today’s social network analysis in its history, and recognises how its current features have been shaped over time under the influence of multiple disciplines, from sociology to social psychology, social anthropology and, more recently, statistics and computer science.

Then Prell moves on to the tools and techniques of social network analysis, and does so with remarkable clarity, unveiling the experience of the teacher behind the hand of the writer: she knows when to warn the newcomer that some principle is not to be overlooked, for example, or that two apparently similar concepts have in fact very different implications. Although she uses some network visualisations in support of her argument, it becomes progressively clear that they are insufficient to come to any solid conclusions about social structures: metrics and indicators need to be calculated, and related to social science theories for interpretation.

Interestingly, Prell organises her topics in a multi-level system: from the individual actor to the dyad and triad, the subgroup, and finally, the entire-network levels. Because the actor level is the most common and familiar in social science, guiding the reader progressively to more and more complex social structures serves a deliberate pedagogical intent. It also clarifies the contribution of social network analysis at each level, and how the different levels nest into, and interact with, one another.

While focusing on theory and methods, the book is also intended as a practical manual for UCINET, one of the most established software packages for social network analysis, and NetDraw, the visualisation application that accompanies it. The reader can easily switch from the theoretical or mathematical presentation of a formula to instructions on how to practically compute it or represent it graphically, using training data sets available with UCINET. In this way, the author firmly grounds her analysis in the tradition of social network analysis and its history, which long preceded the advent of internet and social media: indeed, UCINET has grown within the field, accompanying its evolution and progressively incorporating its findings and advances. Thus, Prell reaffirms the primacy of UCINET against newer software packages, often designed to offer fancy visualisations or to handle the very large network datasets that come from the internet. She takes the risk, though, that the book may become obsolete if UCINET evolves in ways that no longer fit the book’s content or if, in today’s increasing competition, it is superseded by some rival software.

The last part of the book opens to recent developments which are rarely dealt with in introductory materials, particularly the statistical modelling of networks. While descriptive statistical indicators such as mean and standard deviation can be easily computed on network data, regression and hypothesis-testing raise additional challenges because standard independence assumptions do not hold: a network tie between actors A and B may depend on other ties that A and B have, for example because they are both tied to a common acquaintance C. The quadratic assignment procedure, exponential random graph and stochastic actor-based families of models, commonly known as QAP, ERGM and Siena, represent the state-of-the-art in this area. These models are technically complex and may seem intimidating to the less quantitatively versed readers, but Prell achieves the difficult task of presenting their basic features in a simple and concise manner, step by step, and with emphasis on the inner logic rather than mathematical subtleties. She also provides a brief introduction to the StOCNET package used to implement ERGM and Siena, once again facing the problem of software short-livedness, which is even more acute here: indeed, the StOCNET interface is no longer being developed, and more up-to-date versions of the models are now available elsewhere (specifically, in the R statistical environment). Perhaps it would have helped to move these materials to a companion website, easier to update than printed paper.

Overall, the book is a highly recommendable introduction to the field, remarkable for its clarity, accessibility and completeness, with ambitions to fast-forward its readers from beginners’ level up to the point where they can at least appreciate, though perhaps not yet use, advanced models.

Valente’s Social Networks and Health

Valente’s book is an intermediate-level overview of social network theories and methods and their applications, offering an accessible introduction to the newcomer while also serving as a reference for the specialist. It adopts a public health perspective, but because it targets a broad community of epidemiologists who may or may not have prior knowledge of social network analysis, it conducts its discussion at a highly general level that also suits readers in sociology and other disciplines. By the same token, it is also convenient for use in the class, especially at upper undergraduate and at introductory postgraduate levels.

The public health focus of the book does not prevent seamless integration with the perspective of other social sciences and particularly of sociology, whose founding role is fully recognised, particularly in the historical chapter with which Valente, like Prell, opens his exposition. Valente’s chapter is briefer than Prell’s, but proposes an interesting hypothesis as to why social network analysis took so long to spread: it was the social science community’s over-emphasis on inferential statistics and, with it, on the individual level of analysis. To emerge, social network analysis required a shift in mindset, from individual attributes to a relational perspective emphasising social context, inter-individual interactions and shared behaviours.

The ‘Measures’ section at the heart of the book presents basic concepts, tools and techniques of social network analysis; they are not specific to public health and constitute a precious resource for other social scientists, including sociologists. Valente ably manages to provide both mathematical formulas and plain-English explanations of their implications, so that even less mathematically versed readers can follow with minimal effort. Clever, though not overwhelming, use of network visualisations further supports the author’s effort to communicate effectively. Readers will also appreciate his samples of code, which is not tied to any specific software as it comprises both UCINET/NetDraw and standard statistical packages such as Stata. An overview of more advanced statistical models such as ERGM and Siena is provided, though it is more succint than in Prell’s book. The materials provided in the Appendices, particularly a glossary of social network analysis terms, usefully complete the main body of the book. The more quantitatively inclined readers will regret the lack of further mathematical and technical details, though, and Valente himself openly recognises this as a limitation.

Most of the applications showcased in the book derive from six main areas in public health: adolescent substance use, the effects of social support on health outcomes, family planning and sexual behaviour, community health and information diffusion, and, to a lesser extent, inter-organisational collaboration and healthcare provider performance. All these applications have an obvious appeal to medical sociologists and suggest useful ways in which insight into social network structures may illuminate our understanding of disease diffusion, the potential of peer support and self-help groups, and the setting up of health campaigns. If a key goal of the author is to raise the interest of epidemiologists, he also suggests a potentially fruitful collaboration with at least some areas of sociology.

Where Valente really excels is in the sections on ‘Diffusion of innovations’ and ‘Network interventions’, of which he is a recognised specialist. The former illustrates approaches to understand how behaviours spread through a network (for example, how smoking spreads among adolescents in a school-based friendship network), while the latter applies insight from social network analysis to social policy and social work purposes (for example, to design programmes to counter adolescent smoking). Interestingly, the author always starts from a real-world case and builds his argument around the empirical evidence for that case, moving inductively upwards towards more general theoretical assertions. His approach is thus well-suited to the needs of those (even outside the health field) who care most about the practical applicability of social network analysis and its use in policy-making, particularly those who work outside academia and/or in partnership with public-sector organisations, local authorities and NGOs. The author knows that nice visualisations are not enough for these audiences: they want actionable recommendations and concrete results. Most of the effort is devoted to showing that it is worth going the extra mile: learn the techniques of social network analysis, because they are useful.

Overall, the book gently opens the way to cross-disciplinary dialogue and is a most useful tool for medical sociologists, and for social work and social policy specialists in various fields. More generally, it reminds us of the importance of substantive results and practical applications – without which the promises of social network analysis will remain unfulfilled, despite today’s enthusiasm.

Jackson’s Social and Economic Networks

While the other two books aim to strike a balance between the needs of mathematical and non-mathematical readers, Jackson’s one is undoubtedly more demanding. The book draws heavily on mathematics, statistics, and the physics of complex systems in addition to social sciences, and opts for a more theoretical than empirical approach. Jackson is an economist who shares his discipline’s taste for formal methods, and has in mind the high technical standards of the postgraduate economics courses for which his book (with its numerous exercises and imposing list of references) is meant to be used. Yet it is his merit to cover a wide range of topics, without ever becoming too subject-specific, so as to also suit non-economist readers as long as they are sufficiently numerate.

The book is organised around a fundamental tension between random graph and game-theoretic models of networks. These are not just mathematical sophistications but have profound implications: in a nutshell, the former build on probabilistic principles and consider that ties are formed at random, while the latter assume that ties are formed by choice. These are two very different representations of social relationships, supporting different conclusions about network structure and its socio-economic outcomes. Although many sociologists would feel uncomfortable with the probabilistic approach, it might be a useful benchmark against which to measure empirical networks, and an operational simplification in otherwise intractable complex models. The book includes examples for each family of models, illustrating how they can be used; the economic sociologist will be particularly interested in the applications at the end of the book that focus on labour markets, buyer-seller interactions, and supply chain links between companies.

Economists are still new to social networks and remain a minority, albeit a growing one, so that they have to acknowledge – as Jackson duly does – the foundational work of sociologists like James Coleman, Mark Granovetter and Robert Putnam. But their uneasiness transpires with a sense that the sociological core of the field is somewhat vague, has not always been rigorously tested, and may not even be testable at all. Beyond mathematical tools, what is needed are insights that help explain relational behaviours. Jackson explores particularly closely the economic idea that ties have costs and benefits: for example, it is good to have friends, but it also takes time and effort to maintain a friendship (pay visits, stay in touch, do favours, and so on). How people strike a balance between these two competing needs will affect which ties form and endure – all the more so if we believe, as many a sociologist would, that ties are formed by choice rather than randomly. Yet the sociology tradition has often disdained cost-benefit analyses, perhaps because of a certain resistance to associate market-related concepts to a ‘purely social’ relationship like friendship.

On the other hand, there are sociological insights as to other processes through which people form, maintain or delete ties, which economists often overlook or misunderstand. That friendship ties are sometimes reported asymmetrically in network surveys, for example – so that A appears to be friends of B, but not the other way round – does not simply result from flaws in data collection. The sociological literature has investigated this fact under a variety of perspectives, from tie strength (a non-reciprocated tie denoting a ‘weaker’, less frequent or less intense, form of friendship rather than absence of friendship) to social representations (people giving different meaning to friendship) and status issues (those with lower-status over-reporting their friendships); but there is hardly any trace of these discussions in the work of economists, including Jackson. Clearly, these are signs that dialogue between the two disciplines is still difficult and the gap to fill remains large, despite a growing interest in each other’s work about social networks.

Overall, this is a valuable book that raises crucial questions for today’s sociologist interested in social networks. Granted, the discipline has set the ground for others to work and has developed key concepts and theories in social network analysis; but will it be able to stand against the need for greater formal rigour as expressed by economics and with it, the quantitative sciences that also inform Jackson’s work? What can it offer to them today, and, conversely, what can it gain from them?

Discussion and Conclusion

It is a merit of all three books to take social networks very seriously: it is not just visualisation, but analysis; and it is not just about the web, but social networks can describe a variety of social structures. But by the same token, the three books also witness the difficulty of social network approaches to integrate with studies of the internet. Perhaps surprisingly, the two fields have remained separate for long, largely because notions and models that were built for offline networks do not always fit well with the properties of online ones. Only now, some form of integration is slowly starting to see the light (see e.g. Hansen et al., 2010).

The three books all show the importance of the sociological grounding of social network analysis and, in different ways, confirm the importance of inter-disciplinary dialogue. Yet their very juxtaposition also shows how difficult this can be. There is work for sociologists to do in order to make their contribution clear and to address the conceptual and methodological issues that continue to emerge – such as the definition of friendship, which inherits the ambiguities of traditional social network analyses, and to which online friendship adds another layer of complexity today.

The near future will certainly see new developments, and the three books will have contributed to it: by targeting students at various levels, they place themselves as intermediaries using the state-of-the-art of today and the tradition behind it, to inspire newer work. What we see today suggests that sociology can still be part of it tomorrow.