Looking beyond the boundary: Modelling constituency level voting behaviour in the 2024 UK General Election

When modelling areal data, an intrinsic conditional autoregressive (ICAR) component is often used to capture spatial autocorrelation across boundaries. A standard assumption of such models is a global variance parameter, meaning all areas undergo the same degree of spatial smoothing. We propose allowing this variance parameter to change, reflecting different degrees of spatial cohesion in different regions. We demonstrate an application using voting data from the 2024 UK General Election. By comparing the marginal likelihoods of a series of models representing different hypothesised spatial processes, we find that this novel approach is the most plausible model for three of the four largest parties.