The Dynamic Arbitrage-Free Nelson-Siegel Model With GARCH Effect And Treasury Bonds Management Strategies

The term structure of interest rates is not only an important index of a country’s macro economy operation, it influences on the pricing of various financial products in the financial market. At the micro level, the term structure of interest rates adjusts investment and financing decisions, it also establishes benchmark of exchange rates based on Purchasing Power Parity (PPP). The government explicitly proposed "Accelerate the marketization of interest rate and perfect the treasury yield curve" which was included in the core national development strategy. Therefore, it has been a long time searching for better fitting and forecasting evolution model of term structure of interest rates in academic and practical circles.We know the traditional Nelson-Siegel model and the No-Arbitrage Nelson-Siegel model with solid economic foundation are widely used. In this paper, we consider the possible conditional heteroscedasticity in different time and different term to get the Dynamic Arbitrage-Free Nelson-Siegel model with GARCH effect based on the Nelson-Siegel class models.G-AFNS model still have a three-dimensional parameter framework. We use treasury bonds data from January2005to November2012in Exchange Bond Market. We show that our G-AFNS model not only get a high efficiency in fitting exited term structure but also perform well in out-of-sample forecast compared with DNS model and AFNS model. So considering GARCH effect could improve accuracy of forecast and G-AFNS model is suitable for application in treasury bonds market in our country.Next we analyze the treasury bonds management strategies. We mainly focus on the immune bond combination. Consistent with the DNS vector duration, we respectively match level factor duration, slope factor duration and curvature factor duration. As a result, we could eliminate the risk exposure coming from interest rates changes. We adjust immune bonds combination every1month,2months and3months. The result shows that if there are no great changes in interest rates, adjusting portfolio every two months to ensure the immune effect should be scientific and reasonable.