Spatio-Temporal Modeling of second homes dynamics in Corsica

Second home development is an important issue to a housing market and local land use. Understanding second home dynamics at the small area scale will help policymakers to conduct effective interventions to meet the concern of local inhabitants. Conventional approaches do not consider second homes varied across space and over time. Hence, spatial models, such as the Besag, York and Mollie (BYM) model, are useful in terms of offering a feasible way to model both covariates and spatial, temporal dependence. This tool is then applied to analyse the second home ratio at a county level in Corsica, France, over the period 2007-2016. We intend to investigate the impact of spatially referenced covariates and the spatio-temporal dependencies of the second home ratio in each Corsican county. Binomial regression models with spatial and temporal random effects are implemented via Integrated Nested Laplace Approximations pertaining to a Bayesian paradigm. Results show that the model with spatial, temporal random components and a spatiotemporal interaction term produces the most realistic estimates and the best forecast. Then, physical landscape counts, coastal counties, and mountainous counties are positively associated with second home rates, while cultural landscape counts, number of households and interest rates are negatively associated with the second home rates. We also identify the "hot spot" and "cold spot" of the second homes. While the temporally structured component reveals that there was merely a gradual increase on the second home ratio over the past 10 years.