Performance Attribution Method Study Of China’s Bond Portfolio

Portfolio performance attribution is one of the most effective methods to improve portfolio performance. It is one of the most important parts of portfolio performance evaluation system. Performance attribution method is essentially a method which compares the actual performance of the portfolio with that of a market benchmark, and breaks down the excess return into decision-making process corresponding effect to explain the source of the excess return.This paper uses the bond pricing formula as the starting point and applies the calculus methods to decompose bond yields from a mathematical point of view that bond yields mainly come from three aspects, namely, carry return, yield curve return and spread return. Since bond portfolio is a collection of many bonds, the further mathematical formula of decomposition of bond portfolio yield is also obtained. Then, this article further discusses the investment process which is also the commonly used "top-down" investment decision-making process and breaks it down into five major decision factors according to different investment strategy which captures different excess return. Thus, this article finds that the total excess return of China’s bond portfolio relative to benchmark can be attributed to five effects which respectively come from the five major decision factors. These five effects are the carry effect, duration effect, yield curve positioning effect, allocation effect and selection effect.Finally, in order to validate whether the five attribution effects can fully explain the excess return, this article also uses the real data of an investment company to carry out an empirical study. The empirical study results show that:the proposed model is effective when the period is much short such as one trading day, because the five attribution effects can explain more than90%of the excess return. However, when the period is longer such as one month, the proposed model is not effective because the residual is large which means the five attribution effects can only explain some of the excess return. This result also proves that it’s necessary to adopt some smoothing algorithms when examining the attribution results during a longer period.