A social and semantic network analysis of the United States Senate: Focusing on the issue and hyperlink structure
Kim, Jang Hyun
MetadataShow full item record
Although there are numerous studies about congressional decision-making such as roll call votes and their prediction, the communication flows behind such decision-making processes have been overlooked. This dissertation presents hyperlink and issue networks as indices of communication flows among the membership of the 109 th United States Senate and as predictors of senatorial voting. This study used two methodological approaches: social network analysis and semantic network analysis. Social network analysis is a set of research procedures for identifying structures in systems based on the relations among the system's components rather than individual attributes (Barnett, 2001). Semantic network analysis is a network analysis of semantic units in texts (Doerfel & Barnett, 1999). Data on hyperlinks and issue networks were directly captured from each senator's official Web site. Information about seniority, geographical proximity, party affiliation, and roll call votes was acquired from the 109 th United States Senate's official Web site. Political ideology difference data were acquired from Americans for Democratic Action (2007). Campaign funding data was from the Federal Election Commission (2006). Through social network analysis, the major actors in the hyperlink network were found to connect mostly Democrat senators. The centralities of hyperlink networks correlate with the senators' levels of seniority and political ideology. There were 24 shared issues among senators' Web sites. The results showed that Republican senators cover more issues and play a more central role in the issue network. In addition, they deal with so-called 'liberal' issues even more than Democrats. An analysis of the two networks with geographical proximity, party affiliation, political ideology difference, shared campaign funding, and voting similarity networks indicated that physical distance influenced all five of the other networks. Hyperlink networks are significantly correlated with all the networks other than the shared issue network. Party affiliation is positively related to the hyperlink network, and showed a strong relationship with roll call voting and political ideology difference. Political ideology difference also has a strong inverse relationship with roll call voting. Shared campaign funding amount is positively correlated with party affiliation and roll call voting. Roll call voting also showed positive relationships with hyperlinks, geographical proximity, and party membership, with the exception of the issue network. The roll-call voting data were predicted by the five variables with five models. The traditional prediction model composed of party, ideology, and campaign funding explained 69.8% (p = .000) of total variance of the senate roll call votes. Among the five models, the most parsimonious and efficient showed that party, ideology, hyperlink, and campaign funding, predict congressional voting with an increased R 2 (R 2 = .700, p =.000). All of the predictors in the model are significant. The contribution of this study is that the hyperlink network is a good predictor of senators' voting similarity. In sum, hyperlink structure denotes that Democrats use the links for communication purposes. In contrast, Republican senators play a major role in the shared-issue network, which means that Web sites act as a forum for discussion. Also, internet information flow, especially hyperlink networks, should be added to the models explaining senatorial roll call votes. The results indicate that public decision-making is significantly influenced by communication factors.