Network topology effects on the social circle polls

Abstract: Election polls play a critical role in political discussions by probing public opinion and enabling political parties to assess their performance before elections. However, traditional polling methods sometimes fail to predict election outcomes accurately, leading researchers to explore new methodologies. One such approach is the "social circle" question, which asks respondents about the voting preferences of their social contacts. This method leverages collective intelligence and has shown promise in improving predictive accuracy. Nevertheless, the influence of the social network's topology on the effectiveness of social circle polls remains unexplored. In this study, we develop a theoretical framework to analyse how social network structure affects polling accuracy. By simulating voter networks with varying levels of polarisation and connectivity, we assess the performance of both standard and social circle polling methods. Our findings indicate that while social circle polls generally outperform traditional approaches, certain network characteristics can introduce biases and undermine their performances, particularly in polarised situations, which are increasingly frequent in the current political landscape. To address these challenges, we propose a new estimator that combines information from both standard and social circle polls, improving election outcome predictions. We demonstrate the applicability of our method using real-world polling data from the 2016 U.S. presidential election, showcasing its practical utility and providing an estimate of the polarisation level in that society. This work establishes a foundation for enhancing polling methodologies by assessing and integrating network features, which has significant implications for social and political research.