Simultaneously Modelling Conditional Heteroskedasticity and Scale Change

Abstract

This paper proposes a semiparametric approach by introducing a smooth scale function into the standard GARCH model so that conditional heteroskedasticity and scale change in a financial time series can be modelled simultaneously. An estimation procedure combining kernel estimation of the scale function and maximum likelihood estimation of the GARCH parameters is proposed. Asymptotic proper- ties of the kernel estimator are investigated in detail. An iterative plug-in algorithm is developed for selecting the bandwidth. Practical performance of the proposal is illustrated by simulation. The proposal is applied to the daily S&P 500 and DAX 100 returns. It is shown that there are simultaneously significant conditional heteroskedasticity and scale change in these series.