POLITICAL BELIEF SYSTEMS AS NETWORKS OF ATTITUDES: CONCEPTUALIZATIONS AND IMPLICATIONS

This thesis advances the study of political belief systems by integrating network-based approaches from political science, sociology, and psychometrics. The research is grounded in Philip Converse’s seminal work, which conceptualized political belief systems as interconnected networks of attitudes. Despite the foundational importance of this framework, the field has been slow to empirically embrace this networked perspective. This thesis seeks to fill this gap by employing innovative methods to study belief systems. The research is centered around two core questions: First, how can political attitudes be conceptualized as a belief system? This involves refining the construct of a belief system and investigating the implications deriving from considering attitudes as parts of a broader cognitive structure. Second, how can existing network methodologies be integrated to study belief systems? By combining Belief Network Analysis (BNA), Correlational Class Analysis (CCA), and Pairwise Markov Random Fields (PMRFs)—previously isolated within distinct research domains—the thesis promotes cross-fertilization between disciplines. Chapter 2 presents the theoretical and methodological foundations of the belief system literature. Following Converse, I define political belief systems as networks of causally interacting political attitudes. I then discuss why the dominant view on measurement in social sciences — the latent variable model— is limited in studying these constructs. Next, I motivate why network approaches can provide new insights in this field. Finally, I conduct a scoping review of all papers adopting these approaches to study socio-political attitudes. Chapters 3, 4, and 5 present three empirical analyses. Chapter 3 contributes to the growing literature on subjective inequality by applying CCA and Exploratory Graph Analysis (EGA) to survey data from the U.S. and the Netherlands (n = 2,501 and 1,618). These methods uncover distinct cognitive structures in both countries, with CCA identifying groups sharing similar construals and EGA modeling these as belief systems. The analysis shows that the two belief systems of each country have distinct sociodemographic correlates and that attitudes toward inequality are more socially patterned in the U.S. Moreover, we show that belief systems might represent independent factors in determining different levels of support for redistribution. These findings highlight that ignoring the structural organization of attitudes limits our understanding of public opinions on inequality and redistributive policies. Chapter 4 extends the inquiry by examining how multiple political attitudes coalesce to form a political belief system. Using original data from the 2022 Italian general election, this chapter applies three network models to explore variations in belief system structures across different social groups. The results show that political interest increases the interdependence of attitudes, leading to tighter belief systems, while education levels do not have a similar effect. Interestingly, despite ideological differences, voters from Italy’s two main coalitions (left and right) organize their attitudes similarly, whereas Movimento 5 Stelle voters display significantly different belief structures. This chapter addresses a significant gap by showing how distinct political groups organize their support for parties and issues. Chapter 5 introduces a novel approach to Propensity to Vote (PTV) data, conceptualizing PTV as behavioral intentions embedded in a belief system regarding vote choice. Rather than using traditional data reduction techniques, this chapter models PTV data as networks, where items are connected by weighted edges representing their unique shared variance. Using data from the European Election Studies (1989–2019; 28 countries, n = 121,959), this chapter shows that national political contexts influence the structure of PTV networks. Higher levels of party institutionalization are linked to higher network connectivity, while higher affective polarization correlates with the negativity of PTV belief systems (i.e., the higher ratio of negative to positive edges). Furthermore, increased negativity in these networks is associated with higher voter turnout in both EU and national elections, suggesting that PTV belief systems might have important downstream effects on political behavior. In conclusion, this thesis offers both conceptual and methodological dvancements in the study of political belief systems. By integrating network-based approaches, it provides new tools to understand how political attitudes are organized and how they vary across different sociopolitical contexts and sociodemographic segments. The findings underscore the importance of accounting for the structural features of belief systems and highlight the need for further research into their causal dynamics.