Relative impact of stochastic volatility versus stochastic correlation on pricing accuracy

Establish whether, for Monte Carlo pricing of multi-asset basket quanto call options on foreign equity indices with stochastic exchange rates, modeling asset-return volatility as stochastic rather than constant produces a larger improvement in pricing accuracy than modeling inter-asset correlation as stochastic rather than constant. Conduct the analysis within the paper’s framework that considers families of stochastic volatility models (e.g., Heston, 3/2, GARCH-inspired, Bates, GARCH-Jump), stochastic correlation models (e.g., Wright–Fisher, Jacobi, mean-reverting Wright–Fisher, Weibull), stochastic exchange-rate models (e.g., GBM, mean-reverting OU-inspired, exponential Lévy), and common discretization schemes (Euler–Maruyama, Milstein, Runge–Kutta).

Background

The paper compares 60 combinations of stochastic volatility, stochastic correlation, and stochastic exchange-rate models for pricing two- and three-asset basket quanto call options via Monte Carlo simulation. Across tests, configurations with GARCH-Jump stochastic volatility, Weibull stochastic correlation, and OU-style mean-reverting stochastic exchange rates performed best by their accuracy metric.

Based on these empirical findings, the authors hypothesize that making volatility stochastic confers a greater accuracy benefit than making correlation stochastic. They explicitly state this as a conjecture and call for confirmation or refutation within the modeling and simulation framework they paper.