Capturing Four Stylized Facts of Financial Time Series in GARCH and Stochastic Volatility Models
Download as PDF
DOI: 10.23977/ICEMGD2020.051
Corresponding Author
Jiawei He
ABSTRACT
In this paper, four stylized empirical features about the log returns of financial data sets from the point of view in statistical analysis will be described, which are fat-tails, the gain/loss asymmetry, the absence of autocorrelation and volatility clustering. Those properties will be illustrated with three assets from different financial markets. Because these features cannot be captured by any single model, it is necessary to discuss several models which consider kurtosis-autocorrelation combinations and volatilities. The purpose of this paper is to compare three theoretical models, GARCH, ARSV and EGARCH, figuring out how well they can reproduce those stylized features, and examine them through forecasting, which is helpful for modifying and establishing more proper empirical models in order to perform reliable analysis. Finally, it will be pointed out that none of the models dominates the others when comparing their ability to capture four features. However, ARSV can better reproduce the original data set and it can generate smaller value of autocorrelations with larger range of values of parameters.
KEYWORDS
Fat tails, asymmetry-symmetry, autocorrelation, volatility clustering, GARCH, EGARCH, stochastic volatility, ARSV