(An) empirical study on the continuous-time dynamics of KOSPI200 index

Abstract

In this thesis, we estimate and evaluate different types of continuous-time specifications for the dynamics of stock index using efficient method of moment (EMM) with seminonparametric GARCH type auxiliary model augmented with leverage and level effect on conditional variance. We find that while non-normal conditional density of the asset returns and conditionally heteroskedastic variance with leverage effect are significant for KOSPI200 index in the discrete-time representation of SNP, the dynamics the data generating process for the returns of KOSPI200 index does not deviates much from the traditional specification of lognormal process with constant mean and variance on daily level. In addition, extension of the variance specification to constant elasticity of variance model with negative elasticity, and stochastic volatility model with negative relationship between return and volatility innovations improves performance of capturing the features of daily returns.