Robustness and sensitivity analyses for stochastic volatility models under uncertain data structure

TL;DR

This paper evaluates the robustness and sensitivity of stochastic volatility models to market data variations, emphasizing the importance of jump components and long memory effects, using novel bootstrap and Monte Carlo methods on real market data.

Contribution

It introduces a new methodology for robustness analysis of SV models that does not assume parameter independence, validated through empirical data.

Findings

Jump components significantly impact model robustness.

Long memory parameters influence model sensitivity.

Proposed methods outperform traditional sensitivity analysis approaches.

Abstract

In this paper we perform robustness and sensitivity analysis of several continuous-time stochastic volatility (SV) models with respect to the process of market calibration. The analyses should validate the hypothesis on importance of the jump part in the underlying model dynamics. Also an impact of the long memory parameter is measured for the approximative fractional SV model. For the first time, the robustness of calibrated models is measured using bootstrapping methods on market data and Monte-Carlo filtering techniques. In contrast to several other sensitivity analysis approaches for SV models, the newly proposed methodology does not require independence of calibrated parameters - an assumption that is typically not satisfied in practice. Empirical study is performed on data sets of Apple Inc. equity options traded in April and May 2015.