A comparison of Methods for Spatial-Temporal Forecasting With An Application To Real Estate Prices

Svetchnikova, D., Rambaldi, A. N. and Strachan, R. (2008). A comparison of Methods for Spatial-Temporal Forecasting With An Application To Real Estate Prices. In: ESAM08- Proceedings. ESAM08 Markets and Models: Policy Frontiers in the AWH Phillips Tradition, Wellington, NZ, (). 9-11 July 2008.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
Author Svetchnikova, D.
Rambaldi, A. N.
Strachan, R.
Title of paper A comparison of Methods for Spatial-Temporal Forecasting With An Application To Real Estate Prices
Conference name ESAM08 Markets and Models: Policy Frontiers in the AWH Phillips Tradition
Conference location Wellington, NZ
Conference dates 9-11 July 2008
Convener Bollard, A.
Proceedings title ESAM08- Proceedings
Place of Publication NZ
Publisher Economic Society of Australia, NZ Association of Economists
Publication Year 2008
Sub-type Published abstract
Total pages 1
Collection year 2009
Language eng
Abstract/Summary The improvements in the reporting and maintenance of data sets containing spatial and temporal domains as well as powerful computers have opened the way for Spatial-Temporal (ST) models and estimation techniques in many disciplines. In this paper we compare a Spatial Temporal Linear Model (STLM) proposed in the literature to forecast real estate prices to a Spatial Errors Model (SEM) cast in state-space form (SSSEM). We explore in detail the incorporation of the time and spatial information in the estimation of the parameters of both models. We derive analytical expressions that show how the spatial and time information are handled in the estimation of the hedonic parameters in each case. The estimates from the STLM and from the Kalman Filter of the SSSEM account for spatial correlation of contemporaneous and past sales, although the relative weighting of information differs. This is not the case for the Kalman smoothed estimates. The fixed time estimates from STLM are expected to be close to the average of the time-varying estimates produced by the Kalman Filter over the same time period. We illustrate both methods with a sample from Brisbane, Australia for the period 1985-2005. We find the estimation of the STLM model is not computationally feasible for samples larger than 9,000. This is a severe draw back considering the usual size of real estate data sets. A comparison of prediction performance indicates the RMSE of the SSSEM based predictions is considerably lower than those obtained from STLM.
Subjects 1403 Econometrics
9199 Other Economic Framework
E1
Keyword Spatial-temporal
Kalman filter
Real estate
Q-Index Code EX
Q-Index Status Confirmed Code

 
Versions
Version Filter Type
Citation counts: Google Scholar Search Google Scholar
Access Statistics: 122 Abstract Views, 2 File Downloads  -  Detailed Statistics
Created: Sun, 19 Apr 2009, 12:42:12 EST by Kaelene Matts on behalf of School of Economics