We are What We Consume: Predicting Independent Voters’ Voting Preference From Their Media Diet Color

Abstract

Party identification is an important predictor of voting preference, but because a growing percentage of voters do not express any party identification, alternative ways to anticipate voting preferences are required. Partisan slants in voters’ media consumption might offer a relevant proxy. With method triangulation, the current study explores whether media consumption prior to elections can predict voting preferences among independents. Depending on the media outlets adopted by voters and their partisan skew, as detected by Bert machine learning models, the authors calculate an overall partisan slant for each voter’s political information consumption. Data from a nationwide panel survey conducted in Taiwan affirm that their media diet “color” in 2019 can predict independent voters' choices in 2020.