Modeling structural changes in volatility
(2013)
Files
Dufays_thesis.pdf
Open Access - Adobe PDF
- 1.49 MB
Details
- Authors
- Supervisors
- Bauwens, Luc
- Abstract
- (en) Generalized Auto-regressive Conditional Heteroskedastic (GARCH) models with fixed parameters are typically used to model and predict the volatility of financial time series. When estimating such models a common finding is that conditional variances are strongly persistent, especially for long time series. It has been argued that this feature is due to changes in the parameters of the GARCH process, which are overlooked if the model specification imposes fixed parameters. An interesting way of making GARCH models more flexible is enriching them with a dynamic discrete latent state Markov process in such a way that the parameters can switch from one value to another. These models are called Markov-switching (MS) and Change-point (CP) GARCH models. However estimation of MS- or CP-GARCH models is numerically unfeasible either by the method of maximum likelihood or by the Bayesian approach, given the path dependence problem. The thesis develops three different estimation methods that solve the issue. It also details how to choose the optimal number of regimes. The dissertation then explores numerous aspects of abrupt switching models in the context of financial time series. For instance, evidence of structural breaks in volatility processes is empirically observed in each Chapter.
- Affiliations
Citations
Dufays, A. (2013). Modeling structural changes in volatility. https://hdl.handle.net/2078.5/203962