Take me to your Leader: Predicting Political Leadership using Social Network Metrics

The increasing use of the web for sharing opinions and interests has spawned an area of research matching social network metrics - derived from the linking structure of the web - with the predictability of social events (Aizen et al. 2004, Gloor, 2007). Our research addressed the validity of using high-betweenness websites to identify likely winning candidates. We matched national polls’ results with the relative popularity of six candidates for Secretary of the new Italian Democratic Party. We determined web popularity using a degree-of-separation search with the software tool Condor, that computes nodes’ betweeness centrality by the number of links pointing to it (Gloor, 2007). We observed changes in candidates’ relative centrality by identifying the most “talked about” candidates both on the Web and in the Blogosphere. When the search was done on google.com and technorati.com, the results show a negative correlation between the national polls’ results and the betweeness centrality of candidates The same query on Google Blog showed a positive correlation between centrality and the likelihood of winning according to the polls. These findings suggest that a search on the Blogosphere, where knowledge can be shared among experts, might provide a more accurate projection of election outcomes. A possible evolution of this research might be to match political leadership with blog betweenness centrality across different countries and cultures.