Essays on time series analysis
Date
2014
Authors
Shi, Yanlin
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis is a collection of essays on modelling volatility with time series techniques. The first essay addresses the question of modelling structural breaks in the Fractionally Integrated Generalised Autoregressive Conditional Heteroskedasticity (FIGARCH) model. By detecting structural change points via the Markov Regime-Switching (MRS) framework, a two-stage Three-State FIGARCH (3S-FIGARCH) model is proposed. Compared with various existing FIGARCH family models, our empirical results suggest that the 3S-FIGARCH model is preferred in all cases and can potentially provide a more reliable estimate of the long-memory parameter. The second essay examines the confusion between long memory and regime switching in volatility via a set of Monte Carlo simulations. A theoretical proof is provided to show that this confusion is caused by the effects of the smoothing probability from the data-generating process (DGP) of the MRS-GARCH model. To control for these effects, the MRS-FIGARCH model is proposed. By conducting a set of Monte Carlo simulations, we show that the MRS-FIGARCH model can effectively distinguish between the pure FIGARCH and pure MRS-GARCH DGPs. Further, an empirical application suggests that the MRS-FIGARCH can be a widely useful tool for volatility modelling. The third essay empirically studies the relation between public information arrivals and intraday stock return volatility. Motivated by the Mixture of Distribution Hypothesis (MDH) and the study of Veronesi (1999), we fit hourly Standard & Poor's (S&P) 100 stock return data with the MRS-GARCH model to investigate the effect of the quantity and quality of news on stock return volatility in the calm (low volatility) and turbulent (high volatility) states. The effect of news on the persistence and magnitude of volatility depends on the quality of news and the state of stock return volatility. In addition, this effect varies across sectors and firm sizes. The fourth essay analyses the effects of news on the so-called 'idiosyncratic volatility puzzle'. By empirically modelling the stock return data from the Center for Research in Security Prices (CRSP) database from 2000 to 2011, we demonstrate that both the quantity and quality of news can significantly explain the effect of idiosyncratic volatility on excess returns. Specifically, when news effects are appropriately controlled, the average magnitude of this effect can be reduced by roughly 50 per cent.
Description
Keywords
Citation
Collections
Source
Type
Thesis (PhD)
Book Title
Entity type
Access Statement
License Rights
Restricted until
Downloads
File
Description