Research on the Predicting Model of the National Bond Repo Rate in Currency Market in China

The repo rate of the national bond is analyzed, and ARIMA and GARCH models related with the rate are established and the predicting capability of the two models is compared. The result indicts: Model ARIMA(0, 1, 1) has better adaptability for national bond repo rate by the traditional methods of Box & Jenkins on principle of Min Akaike's Information Criterion and Min Schwarz's Bayesian Criterion and the ARIMA model fits nicely for national bond repo rate along with checking the residual through LM series correlative test; Model GARCH(2,3) has better adaptability to adapt GARCH model; ARIMA model's predictive confidence interval is smaller on the 95% CI by the way of predicting day by day for the models constructed. Although forecast volatility of GARCH model is more excellent than that of ARIMA model, we prefer ARIMA model to GARCH model due to ARIMA model's superior predicting capability of the national bond repo rate.