In this article, I present a social network map of United Nations websites and tables of website ranking, based on which websites are most and least popular within the network. At the end, I proved some introductory terms used in social network analysis. Then for anyone interested in the process, I have proved a detailed account of the data collection and mapping process.
In the summer of 2007, I ran a number of tests and experiments to develop my skills in the application of social network analysis to online data. I’ve posted the background for these studies, and below, you will find an overview of my test on the United Nations network.
For my first mid-sized tests, I chose the United Nations family of websites. In July 2007, I obtained a list of United Nations websites from the UN System website (unsystem.org). Selecting 93 websites that had usable URLs, I collected link data by submitting over 8,500 queries to the Yahoo API and then created a matrix of all link relations between each United Nations website. The Yahoo API can sometimes make amazing exaggerations, so to simplify the results and make them more reliable, I converted the number of links from one website to another, to just a binary value showing just one link. In other words, if one website had 50 links to another, I just recorded one link. At this point, I was able to calculate a number of standard social network measures (shown in Table 1), such as the in degree centrality, closeness, between and coreness which are all explained in the box at the end of this posting.
At this point, it was not possible to visualize the network, as there were too many links and the network was too complex. To simplify the network further, I created another network matrix that only included mutually linked United Nations websites. For this successive network matrix, I sought to capture United Nations websites that had the strongest bonds, as manifest by mutually linked websites. To understand this principle, consider that if person A deems B a friend, but B does not feel the same about A, then it would be a mistake to call A and B friends. However, if both A and B consider one another friends, then we would call them friends, and deem their bond stronger than a one way friendship. Just the same, if one website links to another, and this hyperlink is not reciprocated, then it can be argued that the bond between these two websites is weaker than the bond between two websites that link to each other. To represent this concept in the United Nations website network, I only retained websites that shared a two-way, mutual bond. This resulted in a matrix of 87 United Nations websites that shared mutual links. Standard social network analysis metrics were calculated on this network in Table 2.
The network image above shows a visualization of the 87 mutually linked United Nations website, however, you need to view the diagramme in full size to appreciate the relationships. The size of websites is a function of betweeness which is a measure that aims to capture which sites have stronger brokerage roles. Thus the larger websites are more in-between others, and are considered to be better positioned to benefit from their position between disconnected parties. Spatial placement is based on the default multi dimensional scaling algorithm (that I need to look up). Colour represents the results of a core-periphery analysis which identifies the most highly inter-connected websites and the least connected websites that are dependent on the core. In other words, the red websites are like the core in-group, or inner circle, of the United Nations website network, while the blue websites are considered to be on the periphery, and outside the core network where the real action happens. In Table 1, this is called coreness.
Standard social network calculations were produced for the two primary networks, with the sites ranked in order of in degree, so that comparisons may be easily made between tables 1 and 2. Table 1 shows the simple matrix, which is based on all 93 United Nations websites that represent the full network that was analysed. Table 2 shows the mutually-linked matrix of 87 websites, and this network only includes websites that share mutual bonds with other United Nations websites.
I did not undertake this analysis for the purpose of assessment, and consequently, will avoid passing any judgement on the status or ranking of these websites. Interpretation of social network analysis measures and visualizations can be a tricky task and it’s easy to jump to conclusions without considering limitations and potential distortions. For example, it’s easy to assume the well connected nodes in the middle are ‘better’ and represent the influential ‘winners’. Likewise, it’s easy to argue the disconnected nodes are the ‘bad’ or ‘loser’ websites. Perhaps there are times when such judgements are permissible, but the reader must remember that these metrics lack essential information such as the cost of managing each website. If costs were calculated, the central websites may show the least impact per dollar. Regarding context, this is just a map of internal United Nations links, and does not say anything about the websites positioning outside the United Nations network. Sound judgement requires an understanding of each node, and in this case, that means being familiar with each website. For example, high ranking websites may be more a function of marketing efforts rather than links that reflect significant partnerships or substantive relations. Nonetheless, I found the results interesting and thought my former webmaster colleagues at the United Nations may also appreciate the results. I’ll leave judgement up to the reader.
Tables of social network analysis results
Standard Social Network Analysis Terms
The following terms compose basic measures commonly used in social network analysis.
This is a straight measure of the number of links to a node; it is based on the principle that the node with the most links must be the most important. As a limitation, this measure does not factor in the importance of network positioning, nor being connected to other well connected nodes. Consequently, degree centrality is considered a good, but limited measure and normally supplemented by other metrics.
In some cases an actor may have many connections, but may be connected to a disconnected sub-network. This is sort of like being the tallest midget. To overcome this limitation, closeness measures how close one node is to all others. In other words, closeness measures how many hops a node must travel to reach all other nodes, and the node with the highest closeness is the best connected.
In some cases, a person does not need to be well connected to everyone in a network to be influential, nor do they need to be connected to well connected persons; rather, they just need to be positioned between persons who don’t know each other. Betweenness measures a node’s brokerage position, between disconnected nodes. In other words, it measures how well a given node is able to connect disconnected nodes, and benefit from their ignorance. This is the match makers measure.
The inner circle, in-group, the elite–these are all terms that describe groups of individuals who are core to a given network or organization. They’re the ones with access to information and who exert the most influence within networks. Capturing this notion, coreness is a measure that represents a core/periphery analysis which aims to break networks into the ‘in group’ and ‘out group’.
Using images or data If you wish to reuse the images or data on this page, feel free to do so provided you cite the source or provide a hyperlink back to www.onlinesocialmarketing.com. Here is a proposed citation “Cugelman, B. (2008) “The United Nations’ Web Network (July 2007)” Retrieved from: http://www.onlinesocialmarketing.com/online-research/social-networking/united-nations-web-network.htm”