Modelling house price cycles in large metropolitan areas

This thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University LondonThe volatility of house prices can raise systemic risks in the housing market due to the vulnerability of the banking and mortgage sectors to such fluctuations. Moreover, the extreme increases in housing markets have been considered a key feature of the last economic crisis and the run-up to it. Such increases, however, came to a sudden halt immediately before the crisis or directly it began. Despite the recent growth of scholarly work on the role of house price behaviour in economic stability, fundamental questions have yet to be answered: for instance: (i) how far do the nonlinear models outperform the linear models? And how does such nonlinearity explain the asymmetry in the cycle; (ii) what are the main characteristics of house price cycles, and how do they differ over time; and (iii) what kind of policy intervention would stop a real estate boom? This thesis, made up of three empirical essays, aims to take a step forward in answering these questions. The first essay examines whether house prices in large metropolitan areas such as London, New York and Hong Kong follow linear or nonlinear models. The Smooth Transition Autoregressive model was used on a sample of monthly data over the period 1996:1 to 2015:12. The results indicate that linear models are unsuitable for modelling the housing market for the chosen cities. Moreover, strong evidence indicates that real estate prices are largely nonlinear and can well be modelled using a logistic smooth transition model (LSTAR). Estimation results also show different degrees of asymmetry. In particular, the speed of transition between the expansion and contraction of house prices is greater in London than it is in Hong Kong while the speed of transition between boom and bust in New York house prices is the slowest. Further, the forecast results suggest that the LSTAR outdoes the linear model in out-of-sample performance. The second essay investigates the main features of house price cycles in the same major metropolitan areas by providing a reasonable level of discrimination between the cyclical decomposition techniques available for capturing suitable measurements for house price cycles. Through a sample of large cities in several countries, it is shown that the model-based filter is suitable for capturing the main features of house price cycles and the results confirm that these cycles are centred at low frequency. Moreover, there is evidence of substantial variation in the duration and amplitude of these cycles both across cities and over time. The third essay provides evidence that real house prices are significantly affected by financial stability policies. Considering the Hong Kong experience, the results show strong evidence of duration dependences in both the upswing and downswing phases of the cycle. Moreover, the time taken to reach the turning point increases dramatically as the cycle proceeds. The findings also suggest that there is feedback between house price volatility and the policies that affect the housing market. Accordingly, house prices respond with more volatility to any change in the loan to value and lending policy indicators (ignoring the sign of this shock). Finally, the evidence of asymmetry suggests that unanticipated house price increases are more destabilising than unanticipated falls in house prices