Dr Eva Schiffer recently took us (a group of researchers from the International Livestock Research Institute, ILRI) through an introductory session to Social Network Analysis – the concept, the methodologies, its application. Most of us were from the innovation side of the research world where qualitative constructs and processes in systems are easily appreciated. But deep inside (at least for me, and I am sure for most participants) we were anxiously looking for what promised to deliver quantitative bridges through which we could explaining meanings to number-minded die-hards.
The course (and the tools) did not disappoint. All participants were ecstatic about the course, because through SNA we could now analyse systems from a ‘connections’ point of view, especially when considering actors (individuals, groups, organizations, or institutions) and their possible relationships (e.g. information flow, product or service delivery, income payments, influence, etc.). The use of the software programs (VisuaLyzer and UCINet) to graphically display these actor-relationships was invaluable, but even more powerful was quantification of the patterns of relationships. Properties such as size, density and degrees of connectivity, centres of network or actor power and influence, subgroup types and clique or cluster analysis, etc. could be presented in numbers, charts and models
However, when considering how we want to communicate with our professional colleagues we are challenged by definite facts stated clearly about SNA. For one the concept works on very distinct actors rather than populations. A link should come from a clearly defined actor to another. Most published examples seem to only use such systems – Jane is a friend to James, who is friends to John and Mary. But we want to use representative samples for statistical inference about populations? Should SNA only be limited in application to systems like organizations where the actor is an easily observable Mr or Ms So and So? How do we present large numbers of certain actors (farmers, traders, pastoralist communities, or even a population of vectors in disease transmission studies)? How do we present mass media (newspapers, radio)? We wanted to use SNA to analyse innovation systems, e.g. farmers get market information from extension officers; the regulation offices sends out policy messages to its field agents, who share these with traders, etc. Am I missing something here about the potential use of SNA?
The second challenge lies in the fact that actor ties (or links) are indicated in binary measures, i.e. the relationship is only either present (=1) or absent (=0). This may be inadequate if one was to assign scales to the ties for certain types of analysis, e.g. level of ‘influence power’, or percentage of returns earned by one actor from another. For example, farmers can share production information among themselves (how do you present this?!!) but they find links to extension agents for the same more valuable. Or market actors in a value chain are linked by product flow and revenue earnings but returns at each link vary and we want to use SNA to analyse for bottle necks and efficiencies. How can we develop questions that elicit binary responses be framed to show weights of link? How do we use graphical displays and quantified analysis to show such products – after all value chains are in many cases graphical constructs that should tell certain tales?
This is my first post on this log and the challenges I have stated and questions I have posed may be old hat to many out there. If you have already figured that out, please help those of us who are new the wood.
Innovation Works, ILRI