Abstract:
This study is aimed at investigating the term structure of interest rates by using the Vasicek (1977) and the Cox-Ingersoll-Ross (1985) models in three aspects. Firstly, the mean of speed of mean reversion. Secondly, the model fit and forecasting accuracy. Lastly, the ability to make abnormal returns based on an estimated yield curve. The data sample in this study consists of treasury bills and government bonds prices from the Thai Bonds Dealing Center (ThaiBDC) during January 1999 to January 2004. The study reveals that the Vasicek and the CIR models both have a positive mean of speed of mean reversion. In terms of goodness of fit, the CIR model outperforms the Vasicek model. Out of the two models, the CIR model better fits the market data and forecast bond prices due to the residual from the CIR model being lower than that of the Vasicek model. Furthermore, a contrarian trading rule, which buys undervalued assets and sells overvalued assets, was introduced to measure any abnormal returns. The result indicates that the CIR model produces higher abnormal returns than the Vasicek model. Therefore, the term structure of interest rates from the CIR model can be a better benchmark in bond trading than the Vasicek model since the CIR model has a better pricing performance.
