International Evidence on the Predictability of Prices of Securitised Real Estate Assets: Econometric Models versus Neural Networks
Abstract: This paper examines the performance of various statistical models and commonly used financial indicators for forecasting securitised real estate index returns for five European countries: the UK, Belgium, The Netherlands, France and Italy. Within a VAR framework it is demonstrated that the gilt-equity yield ratio is in most cases a better predictor of securitised returns than the term structure or the dividend yield. Predictions obtained from the VAR and other univariate time-series models are compared with the predictions of an artificial neural network model. We find that, whilst no single model is universally superior across all series, accuracy measures and horizons considered, the neural network model is generally able to offer the most accurate predictions for 1-month horizons, while for quarterly and half-yearly forecasting, it is hard to better the random walk with drift. Therefore, investors in securitised real estate assets should be aware that the application of a forecast model without evidence of its predictive performance in a particular market could lead to sub-optimal forecasts.