The Research On The Predictability Of Bond Investment Return

Bond risk premia depict the difference between bond returns and the risk-free rate. At present, the financial system reform is the key to the development of Chinese economy and financial markets, in this context, research on bond risk premia helps to deepen the understand of bond market and yield curve, promote the development of the capital market and know better on the sources of risk. This paper makes theoretical and empirical research on the predictability of China Treasury bond risk premia. analyzes and compares the prediction power among different types of factors from the yield curve itself, the macro economy and short-term financing rate in money market.The main conclusions of this paper are:(1) the level of the yield curve has an average of 70% explanation power to the bond excess returns. Besides, there is another 15% brought by the slope of yield curve, macroeconomic factors and liquidity factors. The slope has more significant additional effect on short-term and long-term excess returns rather than mid-term, while money market factors perform better when predict short-term excess return. (2) For specific explanation variables, some coefficients such as the slope of the yield curve and economic output change along the term of bond excess returns. Significant levels of coefficients with time variation for short-term and long-term excess returns are not the same as well. These differences reflect various risk driving factors on different term of bond excess returns, and the changes in the importance of the explanatory variables in different time stages. (3) The improvement of out-of-sample prediction ability of various types of factors is not consistent with the in-sample explanation ability. The average out-of-sample prediction ability decreased about 20%compared to in-sample results. Yield curve level predicts the best in short-term excess returns, while macroeconomic and money market factors have made very good result for the long-term excess returns with a maximum of 30% improvement.