Using Primary Election Data to Identify Normal Vote Deviations in US Presidential Elections

Keywords: Elections, electoral geography, factor analysis
Session Type: Virtual Poster Abstract
Day: Monday
Session Start / End Time: 2/28/2022 11:20 AM (Eastern Time (US & Canada)) - 2/28/2022 12:40 PM (Eastern Time (US & Canada))
Room: Virtual 45

Abstract

Traditionally, electoral geographers have largely utilized general election data to understand sectional and partisan realignment. As such, there is little work dedicated to the study of primary elections, with limited but notable exceptions coming in the form of guest editorials, and book and atlas chapters. The impetus for this research partially stems from the inability of general election analysis to provide much understanding of the 2016 presidential election. Recent spatio-temporal analyses do not distinguish the 2016 contest from others within the most recent electoral epoch, even though it is considered a historical aberration on many accounts. This research is preliminary to a wider project aimed at exploring how traditional methods of electoral geography may be applied to primary election data. This study specifically explores the utility of using factor analytic techniques to understand partisan shifts through primary outcomes. Traditional T-mode factor analysis is applied to winning candidate proportions in Republican primary election returns in South Carolina. South Carolina was chosen as the study site because of its level of variance in key demographics, and because, as "First in the South," it is one of the few states that is regularly at the beginning of the primary season order. Special scrutiny is given to the 2016 election to determine if the uniqueness that eludes general election research is captured by analyzing one of its primaries. Preliminary results suggest that primary election data may be a useful resource in identifying normal deviations that are not observable in general election datasets.