Spatial Heterogeneity in Real Estate Markets
Which house feature is the most location-dependent? - Marco Lomele, Giovanni Mantovani, Letizia Dimonopoli, and Sofia Villa explore this question in their latest project.
Which house feature is the most location-dependent?
They find that different house open-roles (number of bedrooms, parking spaces, distance from the center…) affect house prices differently across Melbourne’s council areas. In particular, distance has the most variable effect while building area is the most consistent.
In their approach they implement:
- Hierarchical Bayesian Modeling (Python PyMC)
- Mixed-Effects Model with spatial random slopes
- Conditional Autoregressive (CAR) Prior for spatial dependencies
- Inference using No-U-Turn Sampler (NUTS) for robust MCMC estimation
