Between-Person Methods Provide Limited Insight About Within-Person Belief Systems

Belief systems are individual-level phenomena that describe the interrelationships of the political attitudes of a person. However, the modal study of the structure of political ideologies and beliefs uses cross-sectional survey data to estimate what is central to the belief system or the dimensionality of the belief system, aggregating across many people. Cross-sectional data, however, are ill-suited to the task of studying individual-level phenomena because they contain an unobservable mixture of within-person and between-person variation. In this project, we use longitudinal datasets from the Netherlands (representative) and the United States (convenience), spanning between 6 months and 10 years, to we ask whether between-subjects methods can help us understand the within-person structure of belief systems. First, we use Bayesian STARTS models (Ludtke et al., 2018) to assess what type of variance cross-sectional studies are likely tapping into. We find that variability in measures of ideology and political beliefs is primarily due to stable between-person differences, with relatively smaller amounts of variation due to within-person differences. Second, we estimate between-person, within-person, and cross-sectional correlations between all items in our study and find that between-person correlations are larger and in some cases differ in their direction from within-person correlations. Furthermore, cross-sectional correlations are most similar to between-person correlations. Taken together, these findings indicate that the modal study may help describe differences between people, but is ill-suited to tell us about the structure of individuals' belief systems. New methods are necessary for a complete understanding of political belief systems that clarify both between- and within-person processes.