Predictability of HK-REITs returns using artificial neural network

Purpose - The purpose of this paper is to determine if artificial neural network (ANN) works better than linear regression in predicting Hong Kong real estate investment trusts’ (REITs) excess return. Design/methodology/approach - Both ANN and the regression were applied in this study to forecast the Hong Kong REITs’ (HK-REITs) return using the capital asset pricing model and Fama and French’s three-factor models. Each result was further split into annual time series as a measure to investigate the consistency of the performance across time. Findings - ANN had produced a better forecasting results than the regression based on their trading performance. However, the forecasting performance varied across individual REITs and time periods. Practical implications - ANN should be considered for use when one were to attempt forecasting the HK-REITs excess returns. However, the trading performance should be always compared with buy and hold strategy prior to make any investment decisions. Originality/value - This paper tested the predicting power of ANN on the HK-REITs and the consistency of its predicting power.