Comparing volatility forecasting models during the global financial crisis

ABSTRACT

Volatility estimation in financial markets has always been a challenge especially in time of crisis. Once asset prices and investment decisions are highly sensitive to such variable, many different models have been proposed in literature. This article estimates the volatility from a new family of stochastic volatility models called non-Gaussian State Space Models, a subclass of state space models where it is possible to compute exact likelihood. Volatilities of important Asian and Oceanian stock market indexes have been estimated and compared to APARCH model estimates. Results showed that non-Gaussian State Space Models outperformed significantly in both in-sample and forecasting cases.

Keywords:

• APARCH models
• Classical and Bayesian inferences
• Heavy-tail distributions
• Non-Gaussian State Space Models
• Volatility models

MATHEMATICS SUBJECT CLASSIFICATION:

• 62M10