Direct Votes, Deepening Divides? Negative Campaigning and Political Affective Polarization in the 2023 Australian Voice Referendum

Referendums are inherently divisive political events, as they require electorates to make binary decisions on complex and multifaceted issues, and may exert broad societal influence by encouraging affective polarization, or growing mutual distrust and dislike between citizens on opposing partisan lines. Given the global rise in the use of referendums and their expansion into increasingly diverse policy areas, understanding their potential to intensify affective polarization is critical. Affective polarization has been shown to erode democratic resilience by heightening tolerance for undemocratic behavior, deepening winner–loser gaps, and diminishing institutional and social trust. Yet research explicitly linking referendums to affective polarization remains limited, leaving the underlying mechanisms insufficiently theorized and empirically tested. Existing studies identify two partial connections between referendums and affective polarization. First, affectively polarized voters often express stronger support for referendums, particularly when dissatisfied with their electoral representation. Second, referendums on identity or sovereignty issues, such as Brexit or Catalan independence, appear to foster issue-based affective polarization, whereby divisions emerge not solely from partisanship but from deeply held positions on specific conflicts. However, the processes that translate referendum campaigns into polarized affective attitudes remain largely unexplored. One plausible mechanism is negative campaigning by political elites. Previous literature indicates that elite polarization and adversarial electoral communication are key drivers of affective polarization in representative politics. Campaigns that emphasize attacks on rivals rather than promotion of one’s own merits tend to heighten partisan hostility, especially in digital environments where exposure to out-group criticism is frequent. This study addresses this gap by investigating whether and how negative campaigning by political elites through social media advertisements influences political affective polarization in referendum contexts. It focuses on the 2023 Australian Constitutional Referendum on the Aboriginal and Torres Strait Islander Voice (the Voice referendum) as a case study. The referendum offers a suitable setting given documented instances of negative campaigning and the availability of rich post-vote survey data measuring affective evaluations of parties and leaders. Building on the analytical framework of Martin and Nai, the study adopts a sequential mixed methods design. In the first, qualitative stage, the content of Meta and Google advertisements published by political elites during the Voice referendum campaign will be systematically analyzed to identify and categorize different types of negative campaigning, then aggregated at the party level. In the second, quantitative stage, these measures will be linked to the 2023 Australian Constitutional Referendum Study using multiple linear regression models to assess their relationship with affective polarization toward parties and leaders. The research makes three principal contributions. Empirically, it offers one of the first tests of how elite-driven negative campaigning in referendums relates to political affective polarization. Theoretically, it extends existing models to account for voter–party incongruence with regards to the referendum issue, which might shape responses to negative party rhetoric. Methodologically, it integrates primary digital advertisement data with voter survey responses, thereby bridging supply- and demand-side analyses of referendum campaigns and advancing understanding of how modern referendum discourse can reshape affective dynamics in multiparty democracies.