Hate speech on social media is often portrayed as a major driver of political polarization and a serious threat to democratic institutions. Yet a fundamental question remains largely unanswered: does exposure to online hate actually change how people vote? A new study by Alexandre Gori Maia (UNICAMP) and Esther Menezes (São Paulo State Government) offers one of the most comprehensive empirical assessments of this issue, focusing on the 2022 Brazilian presidential election.

Published in Social Science Research (click here for full access), the paper investigates whether hate speech circulating on Twitter (now X) influenced electoral outcomes across Brazilian regions. The study brings together an original dataset combining geolocated tweets with official election results. The authors applied lexicon-based and machine-learning approaches to identify hate speech and employed spatial econometric models to account for geographic spillovers and political clustering in the relationship between hate speech and electoral outcomes.

Contrary to widespread concerns, the authors find no consistent evidence that hate speech altered overall election results. Once past voting behavior, demographic composition, and spatial dependence are considered, hate content directed at candidates does not appear to systematically shift vote shares or increase electoral polarization. These findings suggest that online hate speech largely circulates within ideological “echo chambers,” limiting its capacity to persuade voters outside already aligned political groups.

However, the study uncovers an important and nuanced exception. In localities with a higher concentration of younger voters, hate speech exhibits a counterintuitive effect. Rather than harming the targeted candidate, offensive content is associated with increased electoral support for the victim and reduced support for the opposing candidate. This pattern is consistent with a “boomerang effect,” whereby younger voters—who are more exposed to social media and more open to changing political attitudes—respond to hate speech with empathy and counter-mobilization.

Together, these insights provide a more balanced and evidence-based understanding of the role of hate speech in modern elections, underscoring both the resilience of democratic processes and the need for nuanced policy responses to harmful online content. The paper show that while online hate speech should not be trivialized—given its broader social harms—its direct influence on electoral outcomes may be far more limited and context-dependent than commonly assumed.

Funding acknowledgment: This research was supported by the Brazilian National Council for Scientific and Technological Development (CNPq) and the São Paulo Research Foundation (FAPESP).