Finding The Swing Voter : Methods for Voter Classification

This thesis proposes a theory mapping emotional reactions to political information onto a theory of vote decisionmaking and then further onto measurable survey response. Using on-line processing based in emotion, voters form affective summaries about candidates, which store previous information as an emotional response. The act of voting is treated as a single realization of a probabilistic event, with the relative probabilities of each vote option being an expression of the affective summary. These summaries are expressed as warmness or feeling towards each candidate, which can be captured using the ANES Feeling Thermometer scales. A metric of the difference between the scores given to the Republican and Democratic candidates is used, based in the work of William Mayer. Tlhis metric suffers from significant survey error, but is related to party ID and expressed vote choice, as well as demographic factors and perceived efficacy. Feeling thermometer responses are found to carry meaningful information about a respondent's relationship to the election and candidate preference. Thesis Supervisor: Adam Berinsky Title: Professor of Political Science