Graduate Seminar

Location:  RI 208

Speaker: Walisinghe Madusha De Silva

PhD Student supervised by Andrei Volodin and Arzu Sardarli

Title:  Bayesian Hierarchical Spatial Modeling of Housing Affordability and Ownership Inequality in Canada Using CHSP Microdata (2018-2022)

Abstract:

This research develops a Bayesian hierarchical spatial model using the BYM2 prior structure to examine how multiple-property investor activity influences housing affordability and first-time homebuyer outcomes across Canadian communities. Using Statistics Canada CHSP microdata (RDC-accessed, 2018-2022), the project constructs small-area measures at the Census Subdivision level and applies Integrated Nested Laplace Approximation (INLA) for fast spatial inference. The modeling framework decomposes observed affordability indicators into fixed effects (e.g., investor shares and socio-economic covariates), structured spatial random effects capturing neighbor dependence ("investor contagion"), and unstructured heterogeneity, with Penalized Complexity priors throughout. Key research questions address: (1) spatial patterns of affordability and investor presence, (2) first-time buyer outcomes, (3) spatial spillover effects across communities, and (4) policy scenario projections.

This first seminar presents the complete research design, including motivation, data environment, detailed BYM2 methodology, computational strategy, and planned diagnostics. Expected contributions include novel spatial evidence for investor-focused housing policies and an interactive R Shiny policy dashboard.