On a Semiparametric Stochastic Volatility Model

TL;DR

This paper introduces semiparametric stochastic volatility models using nonparametric techniques and MCMC estimation, improving accuracy in capturing financial market volatility and tail behavior compared to traditional parametric models.

Contribution

It proposes two novel semiparametric SV models with nonparametric error distributions and develops an MCMC method for their estimation, enhancing volatility modeling accuracy.

Findings

Nonparametric models reduce bias and variance in volatility estimates.

Empirical tests on S&P 500 data show improved accuracy over traditional models.

Method offers a flexible alternative for financial risk assessment.

Abstract

This paper presents a novel approach to stochastic volatility (SV) modeling by utilizing nonparametric techniques that enhance our ability to capture the volatility of financial time series data, with a particular emphasis on the non-Gaussian behavior of asset return distributions. Although traditional parametric SV models can be useful, they often suffer from restrictive assumptions regarding errors, which may inadequately represent extreme values and tail behavior in financial returns. To address these limitations, we propose two semiparametric SV models that use data to better approximate error distributions. To facilitate the computation of model parameters, we developed a Markov Chain Monte Carlo (MCMC) method for estimating model parameters and volatility dynamics. Simulations and empirical tests on S&P 500 data indicate that nonparametric models can minimize bias and variance in volatility estimation, providing a more accurate reflection of market expectations about volatility. This methodology serves as a promising alternative to conventional parametric models, improving precision in financial risk assessment and deepening our understanding of the volatility dynamics of financial returns.