Shortmemory, Longmemory and Jump Dynamics in Global短记忆长记忆和跳跃在全球的动态.pptVIP

The Performance of GARCH Models with Short-memory, Long-memory, or Jump Dynamics: Evidence from Global Financial Markets Yaw-Huei Wang National Central University Co-authored with Chih-Chiang Hsu National Central University Motivation Volatility is a key input for both risk management and derivative pricing. ARCH-type models are the most successful and popular framework for describing volatility. In the past two decades, the model has been developed to be more realistic, but more complicated unfortunately. Research Questions Would the more complicated model performs any better for a particular purpose? Conditional distribution: skewed, fat-tailed Memory: short or long If so, can such improved performance be globally valid for different markets? Objectives Develop a nested volatility model based on the EGARCH framework to investigate the performance of (1)short-memory, (2)long-memory, and (3)jump models in terms of (1)model fitting, (2)volatility forecasting, and (3)VaR prediction for 8 relatively large stock markets. The Model The Model Estimation: MLE Maximize the log likelihood function The Model Short-memory: EGARCH (Nelson, 1991) λt = d = 0. Long-memory: FIEGARCH (Bollerslev Mikkelsen, 1996) λt = 0. Jump dynamics: EGARCH-jump (Maheu McCurdy, 2004) d = 0. EGARCH-skewed-t (Hansen, 1994) Measures of Performance Model fitting: Likelihood ratio test: Akaike information criteria (AIC) Volatility forecasting: Mean squared errors (MSEs) Measures of Performance VaR prediction: Likelihood ratio test In practice, a preferred model should have a violation rate which is no greater than the threshold. Data The Datastream stock market indices of US, Japan, the UK, Germany, France, Canada, Italy and Spain . From July 1990 to June 2005. Excluding holiday, there are, on average, about 3,785 observations. Preliminary tests for the absolute values of returns support the existence of long memory in volatility, particularly clear for US and Canada. Empirical