Automatic Mapping of Social Networks of Actors from Text Corpora: Time Series Analysis

Abstract

To test hypotheses about presidential cabinet network centrality and presidential job approval over time and to illustrate automatic social network identification from large volumes of text, this research mined the social networks among the cabinets of Presidents Reagan, G.H.W. Bush, Clinton, and G.W. Bush based on the members’ co-occurrence in news stories. Each administration’s data was sliced into time intervals corresponding to the Gallup presidential approval polls to synchronize the social networks with presidential job approval ratings. It was hypothesized that when the centrality of the president is lower than that of other cabinet members, job approval ratings are higher. This is based on the assumption that news is generally negative and when the president stands above the other cabinet members in network centrality, he or she is more likely to be associated with the negative press coverage in the minds of members of the public. The hypothesis was supported for each of the administrations with the Reagan and G.H.W. Bush having a lag of 1, Clinton a lag of 4, and G.W. Bush a lag of 2. Automatic network analysis of social actors from textual corpora is feasible and enables testing hypotheses over time.