Data-Driven Risk Measurement by SV-GARCH-EVT Model

TL;DR

This paper introduces a novel SV-EVT model that combines stochastic volatility, fat-tailed distributions, and extreme value theory to improve risk measurement and tail risk estimation in financial markets.

Contribution

It develops a new dynamic risk model integrating SV, EVT, and leverage effects, estimated via MCMC, to better capture market tail risks and improve VaR predictions.

Findings

SV-EVT models outperform traditional models in backtesting.

The model effectively captures fat tails and leverage effects.

Enhanced tail risk estimation for financial returns.

Abstract

This paper aims to more effectively manage and mitigate stock market risks by accurately characterizing financial market returns and volatility. We enhance the Stochastic Volatility (SV) model by incorporating fat-tailed distributions and leverage effects, estimating model parameters using Markov Chain Monte Carlo (MCMC) methods. By integrating extreme value theory (EVT) to fit the tail distribution of standard residuals, we develop the SV-EVT-VaR-based dynamic model. Our empirical analysis, using daily S\&P 500 index data and simulated returns, shows that SV-EVT-based models outperform others in backtesting. These models effectively capture the fat-tailed properties of financial returns and the leverage effect, proving superior for out-of-sample data analysis.