Modelling, estimating, simulating: formalizing attitudes towards inequality as a complex network

Attitudes towards inequality are a complex construct formed by perceptions, judgments and beliefs that people hold toward the distribution of resources in a society. Despite this heterogeneity, scholars studied its components in isolation and, as a result, this field is lacking a holistic approach capable of aggregating existing findings. However, some methodological innovations coming from network psychometrics can shed light on its inner structure. In these network models, survey variables are represented as nodes connected by ties reflecting conditional and mutual dependence. In particular, Mixed Graphical Models [MGM] can be fruitfully applied to social sciences data, being ready to investigate relationships between continuous and categorical variables. Moreover, The Causal Attitude Network model [CAN] was explicitly designed to understand attitudes’ complexity, starting from the interactions existing between evaluative reactions. This second model is limited to binary data only but offers the possibility to compute simulations predicting network dynamics. By combining MGM and CAN models, this contribution will investigate the following hypothesis: • The MGM network of attitudes towards inequality will show a small-world structure, where nodes are both highly clustered and connected (H1). • Network Comparison Tests [NCT] will show structural differences between MGM attitudes networks of people with different socioeconomic backgrounds (H2). • A simulation study will show that changes affecting the most central (versus peripheral) nodes are associated with wider changes in the CAN network (H3). To test our hypotheses, we will use Italian data from the 2019 ISSP Module of Social Inequality. To test H1, an MGM network will be built using a broad set of variables tapping perceptions, beliefs, and judgments regarding social inequality in Italy. Later, the sample will be stratified by five socioeconomic variables: household income, educational level, social class, objective social class and subjective social mobility. Hence, ten networks will be compared through NCTs, assessing H2. Finally, a restricted CAN network will be built, and an Ising simulation will record the spillover effects of manipulation attempts targeting both central and peripheral nodes. This paper aims to provide a twofold contribution. First, it constitutes the first attempt to apply network models to the attitude in question. Such an approach can stimulate the integration of different findings that are currently missing in this field. Additionally, attitudes towards inequality have important macro-effects -such as influencing perceived social conflict, redistributive and fiscal public policy arrangements, and stability of political regimes. Therefore, social and political actors could substantially benefit from knowing the inner structure and the transformative dynamics of this articulated construct.