APPLICATION OF THE TRANSFER FUNCTION APPROACH TO HOUSE PRICE MOVEMENT IN SINGAPORE

Abstract

Currently there exists a sizable number of approaches to the problem of forecasting private residential house prices, the most popular method being the econometric simultaneous equation approach as seen from the plethora of work done in this area in developed countries. Data inadequacies, however, rule out such an approach for Singapore; one alternative is to relate the house price index to a smaller manageable subset of variables. For this reason, the multiple time series approach is particularly attractive. We proceed along this path by considering linking the house price index to both the prime rate (taken as a proxy for the non-standard mortgage rate) and the Gross Domestic Product, both of which are believed to affect housing demand directly, although the timing and sometimes reverse movements do raise a number of questions. Since feedback is unlikely to occur, the difficult task of building multiple or multivariate time series models is greatly simplified into constructing a transfer function or dynamic regression model. Although the idea of system closure by restricting to fewer subsets may be intuitively unappealing, it must be borne in mind that simultaneous equation models may also suffer from this drawback by assuming grossly simplified error structures. Reverse movements are usually difficult to detect in these models since lag structures are often specified in a rudimentary fashion as a consequence of the general weakness of economic theory to predict proper lag structures. If one accepts predictive power as the criterion test for discrminating among models, the transfer function model is not easily beaten. This is not to suggest that the model provides a theoretical framework for economic analysis; rather, it simply reflects the naiveness in which the housing market is often looked at.