Scaling, Clustering and Dynamics of Volatility in Financial Time Series
Author | : Baosheng Yuan |
Publisher | : |
Total Pages | : 225 |
Release | : 2008 |
ISBN-10 | : OCLC:1290311311 |
ISBN-13 | : |
Rating | : 4/5 (11 Downloads) |
Book excerpt: This thesis investigates volatility clustering, scaling and dynamics in financial series of asset returns and studies the underlying mechanism. We propose a direct measure of volatility clustering based on the conditional probability distribution (CPD) of the returns given the return in the previous time interval. We found that the CPDs of returns in real financial time series exhibits universal scaling, characterized by a collapse of the CPDs (of different time lags and of different returns in the previous interval) into to a universal curve exhibiting a power-law tail with an exponent of amp;−4. We construct a simple phenomenological model to explain the emergence of VC and the associated volatility scaling. We also study agent-based models of financial markets, and explore the impact of dynamical risk aversion (DRA) of heterogeneous agents on the price fluctuations. We found that the DRA is the primary driving force responsible for excess price fluctuations and the associated volatility clustering. Both our models (phenomenological model and agent-based model) are able to generate time series that reproduces stylized facts of the market data on different time scales. We have also studied general herding behavior often exhibited in financial markets in the context of an evolutionary Minority Game. We discovered a general mechanism for the transition from segregation into opposing groups to clustering towards cautious behavior.