An empirical study of the Chinese short-term interest rate : a comparison of the predictive power of rival one-factor models

Abstract

The thesis uses the one-factor models proposed by Chan, Karolyi, Longstaff, and Sanders (CKLS, 1992) to study the short-term interest rate in China. Eight stochastic models of the short-term interest rate were estimated with Generalised Method of Moments (GMM). All models examined allow the conditional mean and conditional variance to be functions of the current short-term interest rate. According to the investigation of the Chinese one-month inter bank loan rate, strong evidence was found for a mean-reverting feature in the short-term interest yield curve. The volatility of China's one month interbank loan rate was found to be positively correlated with the level of interest rates, but not strongly so. What is more, results from diagnostic tests suggest that the CKLS model, the CIR SR model, and the Brennan-Schwartz model are correctly specified to model the Chinese short term interest rate, so that these three models seem able to capture sufficiently the dynamics of the Chinese short rates. Finally, Theil's U2 statistics and the Diebold and Mariano test (DM test) were applied to explicitly evaluate the predictive power of the single factor models. For China's short-term interest rate the research found that all the one-factor models have similar forecasting ability in terms of predicting China's short term interest rate.