ResIN: A New Method to Analyze Socio-political Attitude Systems

Abstract Investigating attitude systems is of key interest to many social scientists. However, “classic” approaches to studying such systems like linear scaling or dimensionality reduction risk hiding useful information and may artificially narrow the scope of explanatory theories. As an alternative, scholars have recently turned to Belief Network Analysis (BNA), which formalizes mass attitudes as sets of statistically interconnected nodes. However, BNA inherits many of the assumptions typical of latent variable methods which can preclude explorations of more complex phenomena, such as inter-group asymmetries and non-linearities. This chapter discusses Response Item Networks (ResIN)—an extension of BNA which is detached from several assumptions inherent in latent variable models. In particular, we demonstrate how ResIN can detect attitudinal asymmetries without “forcing” specific (linear or monotonic) dependencies onto source data. In ResIN, structural patterns instead emerge organically from co-endorsement patterns. We use simulations and investigate real-world survey data to highlight how ResIN provides parsimonious, yet realistic reproductions of diverse attitude structures in cases that are difficult to handle both for latent variable models and “classic” BNA. We further discuss unitary, bi-nary, and trinary attitude structures based on survey responses across three European countries as well as non-linearities within the US liberal-conservative spectrum.