Modeling and Forecasting for Realized Volatility of Warrants Based on a Threshold Multiplicative Error Model

摘要

In this paper, we extended the Multiplicative Error Model introduced by Engle (2002): a nonnegative valued process is seen as the product of a scale factor which follows a GARCH type specification and a unit means innovation process. By allowing the model change by threshold space, we have introduced a new kind of Threshold Mixture Multiplicative Error Model (T-MEM). Furthermore, the new model was applied to the daily realized volatility series of domestic warrants, which is a unbiased, super-consistent and efficient estimate of low-frequency volatility based on the history, ex-post sample variance of high-frequency data. The empirical results show the new model's good fit and excellent short-term forecast performance comparing to the MEM model.