Time Varying Prediction of UK Asset Returns

Abstract

This paper studies the economic significance of stock and bond return predictability in UK market over period February 1991 to May 2002, based on a real-time investment simulation with only publicly available information. We use ten macroeconomic and financial variables as the predicting variables. At each month, one predicting model for stock return and one for bond return are selected according to predefined criterion. Based on the selected forecasting models, the Sharpe ratio strategy and the Tactical Asset Allocation strategy are used to construct an optimal portfolio. These two active strategies are compared with the passive buy-and-hold stock index strategy, taking into account the transaction costs. The results are really encouraging. In the scenarios of zero and low transaction costs, both active strategies generate substantially higher mean returns with much lower standard deviation than the passive buy-and-hold strategy. Moreover, even with high transaction costs, the active strategies are still superior to passive strategy in terms of Sharpe ratio on the portfolios. This study strongly support that the real-time prediction of UK stock and bond returns is economically significant.

Keywords: Market efficiency, Return predictability

JEL Classification: G14, G15