An empirical comparison of alternative models of short interest rate with repo rate in the inter-bank market of China

With daily data of seven-day repo rate in the inter-bank market of China from June 1999 to July 2003, conditional density of the seven-day repo rate is estimated by the SNP estimator. The density shows obvious heteroskedasticity and no-normality. By the EMM estimator, a number of well-known one-factor continuous-time interest rate models are estimated and tested. The models include the Vasicek model, Brennan-Schwartz model, CIR model, CKLS model, CIR0 model that is a generalized CIR model, and CKLS0 model that is a generalized CKLS model. The empirical evidence shows that the CIR0 model, and CKLS0 model fit the interest rate data quite well. The CKLS0 model is the best among them to fit the data. The evidence also reveals that the instantaneous volatility of the short rate is time varying, and consists of two parts: a constant part and a part that varies with current interest level.