When I first got interested in Social Network Analysis, I lived in Bolgatanga, a small town in northern Ghana. As far as I knew, I was the only one of the 50 000 inhabitants of Bolga who had ever heard of the term or was even faintly interested in the concepts. I learned again, what a blessing the internet is – especially for those who live and work at the margins: Even though I had no peers around me that I could ask, I could learn whatever I needed on the internet.
But I also realized that a lot of the writing about SNA is done in a highly abstract expert language which makes it difficult – even for a social scientist like me – to get the first entry point into this system of thought. Because I wanted to share my knowledge with my colleagues in the field, I had to translate the basic SNA concepts into everyday language. I started with the “centrality measures”. They describe the position of individual actors in the network:
- Degree centrality: How many links does one actor have? E.g. in a money network: How many people does this actor give money to and get money from. Because giving and getting money are very different activities, you can differentiate between in-degree (how many actors you get money from) and out-degree (how many actors you give money to).
- Closeness centrality: How close is an actor to all others in the network? This is calculated is by counting all the links (direct and indirect) one actor needs to follow to reach all others in the network. E.g. someone wants to be up to date in terms of gossip. A high closeness centrality in the information network means that he can spread rumors to everyone very quickly and that he will hear what others have to say immediately. Closeness is a measure of “access”.
- Betweenness centrality: How often does one actor lie on the shortest path between two others? This measure is related to “control” because someone who sits between two others who want to exchange e.g. information or money, can change the story, keep or delay the financial transfer and is informed about everything that happens between these two actors.
- Eigenvector Centrality: How many links do the actors have that someone is linked to? Just counting links (as in “degree centrality”) might not be enough, because it does make a difference whether the actors you are linked to are themselves all dead ends or well connected. The spread of HIV/AIDS is a good example why eigenvector centrality matters. Someone can be absolutely faithful to their partner (degree centrality of 1 in the sexual contacts network); if the partner engages with a lot of other partners (high degree centrality), the faithful one has a high eigenvector centrality (being linked to a “well-connected” actor) and a high risk of infection.
While degree centrality is easily determined by counting the links an actor has, the other centralities require more complicated formulas. Luckily for those who want to understand the network parameters without becoming mathematicians or programmers, there are software solutions to calculate centralities and other network parameters. I like “Visualyzer” because of its user-friendliness, the most commonly used application seems to be UCINet, which is great for the more complex operations; both of them offer free trial versions to download. A basic text that goes into much more detail about network parameters is: Hanneman’s “Introduction to Social Network Methods” (http://faculty.ucr.edu/~hanneman/).
Filed under: theoretical considerations |