Understanding intracity housing market dynamics: a state-space model with Bayesian nonparametric clustering approach

Understanding the intracity heterogeneities in housing market dynamics across microgeographic areas is important but challenging due to infrequent transactions. Unlike traditional methods that use trend-based clustering to improve the accuracy of local housing price and rent indices, we propose a novel hybrid model that combines the state-space model and the Bayesian nonparametric clustering approach to cluster neighbourhoods according to their temporal price volatility. We show that our methods improve the performance of traditional methods by 10-40%, using over 889,428 housing transactions in Singapore between 2006 and 2018. We also demonstrate a practical application of our method – monitoring neighbourhoods’ distinct market reactions to macroeconomic or policy shocks, which has important implications for urban planning and housing investment.