Hierarchical House Price Model Incorporating Geographical and Macroeconomic Factors: Evidence from Australia

This paper proposes a tri-level hierarchical model for house price prediction at Australian suburbs postcode level, integrating dynamics from the national level and the Statistical Areas Level 4 (SA4) level under the Australian Statistical Geography Standard (ASGS). Our study advances house price modelling by introducing a novel framework that integrates risk premium–principal component analysis (RP-PCA), vector autoregressive (VAR) modelling, and an empirical copula approach. Employing RP-PCA to ex- tract SA4-level risk factors and combining these with national-level drivers, we develop a VAR model to capture dynamic relationships. Spatial dependencies in one-step-ahead forecast residuals across sub- urbs are modelled via an empirical copula, further enhancing predictability. Results demonstrate that this geographically conditional multi-factor model, structured hierarchically, increases interpretability and improves short-term forecast accuracy without compromising long-term robustness. Furthermore, this methodology presents a dynamic and granular view of house price trends in Australia. Results highlight key national determinants, including interest rate shifts, gross domestic product growth, and exchange rate variations, particularly in metropolitan urban areas. At the SA4 level, household debt levels, income growth, and population dynamics emerge as critical determinants of price trends, highlighting the interplay of economic and demographic drivers across spatial scales.