Modelování podmíněných rozdělení intra-denních výnosů

Hakala, Michal

Time-series modeling of conditional distributions of intraday returns is of great importance to financial professionals and academic researchers. This work contributes to a methodological and empirical body of knowledge on conditional distributions of intraday asset returns. In Chapter 1, we study distributional diurnal patterns. We propose a new semi-parametric modeling framework for capturing distributional diurnal patterns inspired by traditional seasonal adjustment methods compatible with common models in the literature. Capturing distributional diurnal patterns substantially improves forecast precision. In Chapter 2, we study clustered commonality in the stock market to improve intraday volatility forecasts. Using sectors, industries, and data-driven clusters, we extend the heterogeneous autoregressive (HAR) model with relevant groups or clusters of stocks chosen by regularization methods. In Chapter 3, we propose a model with a novel distribution based on Tukey's H transformation to study conditional kurtosis of intraday returns and spillovers of extreme shocks in the stock market. Empirical applications are based on intraday large-cap US-listed stock returns.

guest :: login Digital Repository
		Search		Submit		Help		About