Modelling and prediction of financial returns using locally stationary time series models

点击翻译按钮获取中文摘要

Modelling the returns of financial instruments is essential to modern risk management. Most of currently used models rely heavily on the assumption of stationarity. The significant changes in the economic and financial environment have induced researchers to question whether this hypothesis is a realistic once. The current project focusses on developing time series models under the less restrictive assumption of local stationarity. We consider univariate and multivariate multiplicative models of a deterministic, timechanging volatility sequence with a sequence of independent and identically distributed innovations with an asymmetric heavy-tailed distribution. In the multivariate set-up, we intend to address the modelling of high dimensional vectors of returns by developing a factor-based methodology of dimension reduction in the locally stationary framework.