By switching to dark mode you can reduce the energy consumption of our digital service.

A review of applications of Social Network Analysis

This article reviews the applications of Social Network Analysis in eight diverse studies; details of the methods, questions asked, findings and tools used are given.
Multiple Authors

This section is the main body of the article (citation below) which is summarised here.

Case 1: Network governance and regional resilience to climate change: empirical evidence from mountain tourism communities in the Swiss Gotthard region

T. Luthe, R Wyss, and M. Schuckert, 2012. Network governance and regional resilience to climate change: empirical evidence from mountain tourism communities in the Swiss Gotthard region, Regional Environmental Change 12(4) 839-854 DOI: 10.1007/s10113-012-0294-5

Method:

All stakeholders, as identified by marketing organisations, involved in the tourism supply chain in three locations, as well as at the regional level in Gotthard were contacted. Using a postal survey, the study gained a response rate of 42% (71 people), this is rather low for a method which aims for complete network information rather than sampling (the authors point out the problems with limited representation, such as validation of links) however a large variety of actors participated; six different sectors were included in addition to the public sector. Actors with high ‘betweeness’ centrality (also identified as ‘gatekeepers’ of the network) also participated in follow-up interviews to assess representativeness of the sample.

Network measures were calculated (density, centralization) as well as measures to understand the existence of subgroups (modularity) and role of individuals (centrality, brokerage). A transformation (deleting ties not lying between central actors) was applied to assess the subgroup characteristics of the central ‘core’. Likewise, by deleting the same central actors (and their ties), the connectivity within the periphery was investigated. The network was analysed by location and by sector to show integration and brokerage function of individuals. The authors also investigated vertical connectivity between regional and municipal levels.

Question: Participants were asked whom of the other actors (focusing on the institution, not on individuals) listed they have been in contact with in a professional context.

Findings:

Main findings were low density of the network (7.2%), low clustering, and high centralization /core-periphary structure. Investigation of modularity (more dense links among group members than externally) of sectors and locations characterised ‘gastronomy’ and ‘activities’ as possible subgroups. High network density was found within each of the 3 locations compared to overall density indicating spatial integration, and strong sectoral integration among most sectors except entertainment (lacking connections with 2 other sectors). Concering individual roles, the ten most central actors (for each of 3 centrality measures) are listed and discussed. Various sectors are represented in these lists, and this is taken as an indication of resilience of the network, although the lack of integration of cableway companies could be a potential risk (more related to preparation to longer-term changes requiring social learning, diversification and innovation rather than to response to immediate shocks). Network analysis is carried out in the framework of climate change adaptation and resilience of social systems.

Tool: Visone and Gephi

Case 2: Understanding household connectivity and resilience in marginal rural communities through social network analysis in the village of Habu, Botswana

Cassidy, L., and G. D. Barnes. 2012. Understanding household connectivity and resilience in marginal rural communities through social network analysis in the village of Habu, Botswana. Ecology and Society 17(4): 11. http://dx.doi.org/10.5751/ES-04963-170411

Method:

Authors surveyed the majority (145 or 80%) of households in the community of Habu, speaking either to the head or another available adult to collect household network and other data. Network data focused on exchanges during stress periods relating to disease in livestock, crops or humans. Centrality of households was calculated, to explore the idea that high degree centrality and high betweeness centrality can enhance resilience of households. They carried out data processing steps to produce clearer networks (removing isolates and pendants) removing those with low betweeness values; also three actors with very highly institutionalised roles (and almost 100% in-degree centrality) were removed for the analysis. Compared to the other SNA studies reviewed this one has the largest dataset (which means that network measures, and other statistics may be more reliable).

SNA measures of connectivity are evaluated for correlations with indicators of resilience (household capitals, other attributes of households such as size, gender and age of head, and other connectivity-related variables such as attendence and membership in community organisaitons). These tests were run both with the overall network (which may be obtained by collapsing the four types of relations into one) and ones constructed on specific types of exchange (labour, food, money and information).

This is a good example of how to combine SNA with other statistical measurement and testing for a very comprehensive analysis. However, concerning resilience, it should be noted that by only getting the perspectives of the household head – authors don’t assess differences in individual resilience among heads and other household members, which may be important.

Question: With which three other households do you exchange information, labour, food, or money in times of stress?

Findings:

Main findings are that male-headed households have significantly higher degree centrality and betweeness centrality, as have households with older heads and with larger size, but that ethnicity and educational level are not highly correlated with network centrality. Looking at specific types of relations, labour and food networks were found to be denser than information and money; all networks are are strongly correlated with measures of resilience except the information network which is significantly correlated. The study found differentials among households, and by investigating the profiles of particular weakly connected/ strongly connected nodes provided further insight related with network position. The authors suggest that high degree centrality can enhance resilience of households (by providing redundancy / availability of alternatives) and by facilitating social learning. However they also found that betweeness centrality was the measure that was most often associated with resilience — suggesting that indirect connectivity can also add to resilience and social learning as it provides broader connectivity. Tool: UCINET and SPSS

Case 3: Building ties: social capital network analysis of a forest community in a biosphere reserve in Chiapas, Mexico

L.Rico García-Amado, M. Ruiz Pérez, I. Iniesta-Arandia, G. Dahringer, F. Reyes, and S. Barrasa. (2012) Building ties: social capital network analysis of a forest community in a biosphere reserve in Chiapas, Mexico. Ecology and Society 17(3): 3. http://dx.doi.org/10.5751/ES-04855-170303

Method:

Authors surveyed household heads belonging to a forest community in the ejido Sierra Morena, Chiapas, Mexico. Ejidos are a form of of common property, that provide a way of controlling land access based on inheritence. The village is composed of ejidatario (25) and poblador (7) households, with only the former able to own land, and the authors surveyed all heads. This is a rare example of having highly complete network data, and the authors mention very good researcher-interviewee relations there (based on previous long-term field visits). The study is also characterised by a well-defined boundary.

Directional data were collected for five different labour-related networks, 4 specific ones: coffee, palm, ecotourism and authorities and one general network. It is not reported how these specific types were suggested. Network centralization measures (categorical core/periphery analysis) were calculated on the set of all relations, as was network hierarchy (Krackhardt algorithm). Its sub-structures (Girvan-Newman clustering and factions) were identified. The authors then used specific network information to assign each node to the network in which it had highest indegree and reveal grouping activites in the all-relations network. For each specific network statistical tests for correlations between degree-centrality and group affiliation (tenure status and income category) were performed.

The second part of the paper concentrates in more detail on coffee networks and how they have evolved over five timeperiods since 2000, and this becomes a very interesting analysis combined with other fieldwork data, and the same statistical tests were made for coffee networks as for specific networks. Methodology, there could be a problem with recall for past historical data, and it is not clear how to cater for actors leaving or coming into the system (the authors analysed the same set of nodes over 2000 to 2011) which might lead to incomplete data. However, because of the high quality of the Sierra Morena data and access to interview respondents for cross-validation, this can be assumed not to be a problem here.

Question: To whom do you relate/work with for different productive activities? Which coffee groups did you belong to in the past (2000/2004/2008/2010)?

Findings:

Authors found evidence for network centralisation and heirarchy in the all-relations network and node assignment showed the presence of a separate ecotourism group/cluster (explained by historical rivalries). Differences among networks, for example low transitivity of coffee (2.2% – emphasising open network) and high outdegree centralization of palm (50% – emphasising within-group bonding ties) showed the differences in organisation and were well explained through examining the marketing and regulatory factors. In this way the paper demonstrates that SNA can help understand embeddedness and power structures. Statistical tests showed ejidatarios (group) having significantly higher indegree but not outdegree. Some misleading terminology was used eg. ‘factors determining the position in the network’ ‘income and tenure affected particularly the indegree’ when correlations relations were found. In terms of consequences of the organisation of social networks, the environmental sustainability of production and its resilience were aluded to rather than examined in depth.

Tool: UCINET 6.0 and NetDraw

Case 4: A social network approach to analyzing water governance: The case of the Mkindo catchment, Tanzania

C. Stein, H. Ernstson, J. Barron (2011) A social network approach to analyzing water governance: The case of the Mkindo catchment, Tanzania, J. Phys. Chem. Earth (2011)

Method:

A pre-study is used to produce a recall list – a complete list of all the actors involved. The pre-study involves asking members of a group of informants with long experience to name actors that could potentially influence management of the resource. In Stein et al. study spokespersons of the organisations on the recall list were interviewed and asked to mark their “regular/long-term” relations to other actors on the recall list, for each of the three types: i) funding, ii) information and knowledge exchange, and iii) collaboration. The third type of tie was only counted if reciprocal. It was considered infeasible to interview/visit all actors involved in water management and use, so 4 communities were selected. Then two criteria were used to determine actors’ inclusion/exclusion, ie. to draw the boundary of the SN. The first criterion was attributional, the second criterion was a relational one – that they were mentioned more than twice by respondents as actors to which they had relations. Hence respondents could define the network boundaries – an approach known as ‘expanding selection’. In fact only 2 new actors met this second criteria.

For this paper, the authors only analysed collaborative (reciprocal) ties, investigating centrality of actors as well as centralization of the overall network, density and sub-groups (Newman-Girvan algorithm). Question: What are your “regular/long-term” relations to other actors on the recall list, for each of the three types: i)funding ii)information and knowledge exchange iii)collaboration.

Findings and lessons learned:

Findings were presented at multiple scales of interaction and considered direct and indirect actors separately and together. At the horizontal/local scale ‘direct’ actors (e.g CBOs) are sparsely connected, however addition of the (indirect actors) village leadership improves this picture. Subgroups were found to follow community boundaries with 4 clusters. At the vertical scale, although actors with formal role in governance were found to be central connectors overall, they were not well-connected to local, direct users. This highlights the top-down nature of governance that fails to consider informal structures which may be able to inform and coordinate improved governance.

Tool: UCINet software was used

Case 5: Innovation and diffusion of sustainable agricultural water resource management in a changing climate: A Case study in Northeast Thailand

M. Mikhail, A. Fencl, S. Naruchaikusol and E. Kemp-Benedict (2010) Innovation and diffusion of sustainable agricultural water resource management in a changing climate: A Case study in Northeast Thailand Stockholm Environment Institute Project Report

Method:

The authors apply SNA to study the diffusion of water resources management innovations, specifically the technological improvements in practices of agricultural smallholders, that aim to address climate adaptation. They assessed both the farmers networks and the institutional networks by means of interview-based surveys and they identified four sub-groups of respondents: innovators (without specifying how these were determined), adopters, failures and non-participants. However, the authors were working with very a small number of respondents.

Questions asked: Who in the past 6 months have you talked to about important agricultural matters? How is this person related to you? Are you asking them for advice (a), or are they asking you (b), or is it more of a discussion c) ? Rank them in order of the best advice (1 is best).

Findings and lessons learned:

This is a good example showing that quite a lot can still be said even when working with partial data sets. The authors faced a problem of finding enough participants to interview. Snowballing technique, beginning with a few recommended ‘innovators’ provided a small dataset. Ego-centric data shows how individuals are embedded within their own networks (as they perceive them): analysis of this can reveal some of the diversity among individuals. Therefore, rather than looking at whole network measures, the authors looked at the measures for individual actors, i.e. comparing network statistics for ‘innovators’ in the communities, with the rest of the respondents (non-innovators) and using t-test for significant deviations.

They found a contrast between the village where adaptation has been programmed and the village where it has not taken place – only in the former is betweenness centrality significantly higher among innovators. But the authors do not draw strong conclusions about findings mentioning the difficulties with collecting reliable network data.

Tool: ORA software was used

Case 6: Power Asymmetries in Small-Scale Fisheries: a Barrier to Governance Transformability?

B. Crona and O. Bodin (2010) Power Asymmetries in Small-Scale Fisheries: a Barrier to Governance Transformability? Special Feature on SNA in Natural Resource Governance in Ecology and Society

Method:

Household heads were interviewed in a rural village community highly dependent on fishing. 171 out of 206 were interviewed to collect information on networks used for exchange of gear (of high value such as boats or nets) and knowledge relating to the state and extraction of natural resources. Recall method was used (meaning not using predefined list) to obtain directed relations. In this village there are few lenders and many fishers borrowing – the assumption is that “gear ownership implies a form of power over those dependent on using it” and those with high in-degree (relations in which it is the target) have high power (many others turn to them to borrow gear).

Two networks are constructed involving the same actors : lending network and local ecological knowledge network (LEK) and the authors test to what extent are the central agents in lending network also central agents in the LEK (ie. ‘opinion leaders’) by means of correlation. Blockmodelling method to obtain categories of stakeholder reportedly resulted in a good match (25 deviations) from the three pre-specified categories of multi-source clients, (single source) clients, and owners. (Pajek software used) Questions asked: Lending network: Q: Is there any person(s) on whom you depend, or who depend on you to carry out your (their) occupation? Knowledge network: Q1: If you noticed changes in the natural environment who would you discuss this with? Q2: Do you exchange information with anyone which is useful for you to carry out your common occupation?

Findings and lessons learned:

A key finding was correlation between structural position of actors in the power network (constructed based on gear exchange) and the LEK network. That there are individuals central to both networks simultaneously may have implications for transformability of governance of the fisheries.

Tool: Not specified

Case 7: Can properties of Labor-Exchange Networks Explain the Resilience of Swidden Agriculture?

S Downey (2010) Can properties of Labor-Exchange Networks Explain the Resilience of Swidden Agriculture? Special Feature on SNA in Natural Resource Governance in Ecology and Society

Method:

This study of traditional Swidden agricultural societies combines SNA of labor exchange relations with the concept of adaptive cycles. It examines the carrying capacity of Swidden communities and their resilience. In 2007-2008 a household-level social-network survey was conducted in 5 villages in Toledo District, Belize. Question “If you needed 10 men to help you chop or plant a field, who would you ask?” The article reports that 64% reported talking to others about where to locate their field (did not give the exact phrasing of this question) – basis of statement “survey results show that environmental knowledge is learned and shared as farmers consult each other”. The author then constructs networks based on the indicated directed ties. Network recipricosity, hierarchy (Krackhardt 1994) and efficiency were measured. One limitation is that the small size of surveyed farmers in each village (8, 16, 17, 20, 19, 47) and the total number of households does not seem large enough to do statistical testing.

Question: “If you needed 10 men to help you chop or plant a field, who would you ask?”

Finding:

Social networks become more hierarchical and more efficient with age of village (resulting in smaller work group sizes, and placing a constraint on overuse of land resources). [Could this also be affected by village sizes?]. It is argued that adaptation of the network of labour exchange has a role in protecting the commons and making the Swidden system resilient.

Tool used: UCINet

Case 8: Adaptive co-management Networks: a Comparative Analysis of Two Fishery Conservation Areas in Sweden

A. Sandstrom and C. Rova (2010) Adaptive co-management Networks: a Comparative Analysis of Two Fishery Conservation Areas in Sweden. Special Feature on SNA in Natural Resource Governance in Ecology and Society

Method:

Two fisheries in regions in middle and inland Sweden were compared in terms of the co-management of respective FCAs (there is little contextual infomation on the two FCAs, detailing how or why these were chosen). The authors examied network structure among rule-making actors (persons in management roles) in these two networks (A and B). Concepts of network closure and network heterogeneity are examined in terms of how they relate to adaptability. Closure is measured by network centralization (how star-like the structure is) and network density, whereas heterogeneity is measured by actor diversity (non structural) and cross boundary exchange (percentage of ties connecting actors from different organisations). Burt’s (2000) concept of (bridges over) structural holes cannot be captured with the methodology, which considers only reciprocated links (the argument is that strong ties provide more accurate results). Heterogeneity is hypothesised to enhance available resources such as ecological knowledge, in a similar way to weak ties.

The network consists of actors (of different types) involved in rule-making (regular discussion and communications concerning the FCA. Initial interviews to identify actors, with snowballing technique to allow respondents to name others ‘involved in management’ and to nominate as ‘important’ (ie. to be interviewed). To these ‘involved’ (24 individuals in each case) a survey was administered to ask who they would usually talk to about the goals, rules and routines of the FCA. Authors note that using a small empirical sample results may not be statistically robust.

The interview data (on rule-making processes of FCA) were used in the comparative analysis of adaptability and the network data for structural analysis of rule-making.

Questions asked:

Question 1 asks who you usually talk to about the goals, rules, and routines of the FCA. The question is followed by a list of names. Your task is to mark the persons to whom you usually talk concerning these issues. Question 2 asks to whom you usually talk concerning the ecological status (i.e., the physical condition of the fish and waters of the FCA)? The question is followed by a list of names. Your task is to indicate the persons to whom you usually talk concerning these issues.

Findings and lessons learned:

The SNA shows that Network A has a more cross-boundary character (heterogeneous) and higher closure (centralization). Small set of respones available.

Tool used: UCINet

Overall conclusion

This review had a methodology focus* on recent SNA work on adaptive governance, adaptation and resilience. It has shown that, firstly, a great deal of thought goes into the initial design of the study methodology. This starts with defining what is an actor (node) and what sort of relation (link) should be considered. We have seen in these five papers many different types of actors and linkages can be considered as components of an SNA study. The choice could be the basis for deliberately describing a more stable network structure – to counteract the criticism that SNA can only provide a static view, for example.

In these papers, whereas study design decisions (such as the drawing of network boundaries) vary depending on research objectives, the free recall and snowballing techniques are commonly used. Network centrality measures as well as degree centrality indices (five out of five papers) are used, but these are combined in interesting ways with a variety of other measures, structural and non-structural, (for example blockmodelling, subgroup/clique analysis). Of course, many other design decisions often come down more to practicality and feasibility issues. Also included are examples of comparative network studies, which can also be used to consider the evolution of social systems in terms of structuring of relations.

Amongst the alternatives mentioned, UCINET is the most commonly used software program for data analysis (three studies). See also the table at the top of the summary section of this article.

*The focus on methodology is due to the authors’ involvement in methodology orientated research projects and project workpackages: MEDIATION and emBRACE .

Please let us know if you have found this review useful or informative or if you have any suggestions about other work that you think we could include in this section. Thanks!

SNA review summary

Add your project

Exchange your climate change adaptation projects and lessons learned with the global community.