Negativity Bias in Political Perception Revisited

• In Event: Challenges in Using Large Language Models

Abstract

For decades, negativity bias, where negative information exerts a stronger influence on political perceptions than positive information, has been considered a cornerstone of voter behavior. But as the political landscape grows increasingly negative, is this bias still as dominant as once believed? This paper explores whether positivity is gaining ground in shaping voter evaluations. Using data from the American National Election Study (ANES) between 1968 and 2008, Study 1 conducts a time-series analysis to examine shifts in voter sentiment. The results indicate that while negative information continues to be a dominant force in shaping candidate evaluations, the influence of positive information has gradually increased, particularly in Democratic candidate assessments. This suggests a diminishing negativity bias and the growing relevance of positivity in modern political perceptions. Recognizing the limitation of uncoded responses in post-2008 ANES data, Study 2 employs a generative AI model to classify open-ended responses from the 2020 ANES. The AI-derived model (ADM) was tested against hand-coded benchmarks, demonstrating moderate accuracy in replicating human coding, especially in straightforward responses. However, challenges remain in coding nuanced or ideologically complex content, highlighting areas for further refinement. The study underscores the viability of AI in scaling qualitative research while acknowledging the need for continued advancements to capture the full depth of voter sentiment. Together, these studies suggest that voters may be more responsive to positive cues in an increasingly negative political environment, pointing to a possible paradigm shift in political communication. The findings carry significant implications for both campaign strategies and scholarly research, highlighting the importance of adapting to evolving voter behavior and leveraging AI for more efficient survey analysis. This paper ultimately calls for a reexamination of the enduring assumptions about negativity bias, proposing that the political landscape may be entering a phase where positivity plays a more influential role.

Authors

• Richard R. Lau, Rutgers University, New Brunswick

• Barea M. Sinno, Rutgers University

• George D. Quinn, Rutgers University