Detecting spatial and temporal house price diffusion in the Netherlands: A Bayesian network approach

Abstract

Following the 2007–08 Global Financial Crisis, there has been a growing research interest on the spatial interrelationships between house prices in many countries. This paper examines the spatio-temporal relationship between house prices in the twelve provinces of the Netherlands using a recently proposed econometric modelling technique called the Bayesian Graphical Vector Autoregression (BG-VAR). This network approach is suitable for analysing the complex spatial interactions between house prices. It enables a data-driven identification of the most dominant provinces where temporal house price shocks may largely diffuse through the housing market. Using temporal house price volatilities for owner-occupied dwellings from 1995Q1 to 2016Q1, the results show evidence of temporal dependence and house price diffusion patterns in distinct sub-periods from different provincial housing sub-markets in the Netherlands. In particular, the results indicate that Noord-Holland was most predominant from 1995Q1 to 2005Q2, while Drenthe became most central in the period 2005Q3–2016Q1.