Bayesian hierarchical age-period-cohort models with time-structured effects: An application to religious voting in the US, 1972–2008

To examine dynamics of political processes using repeated cross-section data, effects of age, cohort, and time period have to be disentangled. I propose a Bayesian dynamic hierarchical model with cohort and period effects modeled as random walk through time. It includes smoothly time-varying effects of covariates, allowing researchers to study changing effects of individual characteristics on political behavior. It provides a flexible functional form estimate of age by integrating a semi-parametric approach in the hierarchical model. I employ this approach to examine religious voting in the United States using repeated cross-sectional surveys from 1972 to 2008. I find starkly differing nonlinear trends of de- and re-alignment among different religious denominations.