Empirical Research On The Yield Curve Of China's Treasury Bond

Treasury bond yield curve describes the relationship between yields and the remaining term of the bonds. Treasury bond is guaranteed by the national credit, and the risk of default is very small. Treasury bond yields can generally be took as the risk-free rate. Treasury bond yields have many important applications in different aspects, it can be used as the benchmark interest rates for the pricing of various financial products; Treasury yields move closely with macroeconomic. and are important reference for monetary policy; Treasury yields can help investors to seize market benefits and manage risk, so it is of great significance to study the Treasury bond yield curve.This thesis focus on the Treasury bond yield curve, in order to find the best term structure model to fit the Treasury bond yield curve:and analyze of the causes of the Treasury bond yield curve furthermore.This thesis use empirical analysis method to study the Treasury bond yield curve. Firstly, we collect the relevant theories and models on term structure, and point out that some term structure models are derived based on the experience of foreign markets, and they may not be suitable for the domestic markets. Secondly, we acquire transaction data from the Shanghai Stock Exchange and the Interbank Market. Then we use nonlinear regression model, NS model, BP network model. RBF network model to fit Treasury bond yield curve. Lastly, we evaluate the fitness from the curve shape, standard error and R square, to select the best model. We further analyze the causes of the features of the curve. The main achievements of this thesis are as following:(1) The transactions data in Shanghai Stock Exchange and the Interbank Market shows the distribution of the Treasury bonds is not reasonable. Medium-term bonds are much more than short-term and long-term bonds. Some Treasury bonds don't have any transactions in consecutive days. The liquidity of the Treasury bonds market is poor.(2) BP neural network model with single hidden layer and double hidden layers are better fitted for our market. The optimal structure for the single hidden layer model is two neurons, the optimal structure or the double hidden layers model contains 3 and 5 neurons respectively.(3) From the static point of view, the Treasury bond yield curve is increasing, the slope of the curve is decreasing. The reason of such characteristics lies in that:Firstly, long-term bonds has less liquid than short-term bonds; Secondly, China's economic growth is slowing down, investors'have lower expectations for the market.(4) From the dynamic point of view, the reasons of the changing shape of the yield curve including:the macroeconomic environment, investors and monetary policy. Where yields move closely with macroeconomic, and are positively associated with GDP, CPI; yields are also influenced by investors, including bank insurance and fund; monetary policy has a significantly impact on yields as well, with the central bank lowered five times the benchmark deposit interest rates in 2015, the yield curve also declined.Combining these findings, we propose two suggestions:Firstly, broaden the issuance of different kinds of bonds, and the issue s periodically; secondly, develop the derivatives market and market maker system to increase market liquidity.The innovation of this thesis lies in two areas:Firstly, the innovative research methods. In this paper, we use a variety of neural network structure to fit the Treasury bond yield curve, and find out the optimal BP network and RBF network structure; Secondly, we use the latest data. In this thesis, the empirical research on the Treasury bond yield curve are based on the latest data of the bond market in 2015, therefore, the findings are more time-efficient.