Keeping Up with the Correlations: Stochastic Spot/Volatility Correlation and Exotic Pricing

Battle Count: 68

Highly relevant for FX derivatives market making and exotic options trading desks. The paper demonstrates that using the standard Heston model for pricing one touch options, knockout options, and volatility swaps leads to systematic mispricing comparable to or exceeding bid/ask spreads. This directly impacts P&L for market makers who accumulate mispriced short positions. The risk reversal beta dynamic is a key factor in FX exotic pricing. The model provides actionable insights for improving pricing accuracy and risk management in liquid FX derivatives markets. However, it is more relevant to derivatives pricing desks than to algorithmic trading or portfolio management.

Vanilla option pricing is straightforward via closed-form characteristic function and FFT (Carr-Madan), similar to standard Heston implementation. However, barrier option pricing requires Monte Carlo simulation with Andersen (2008) QE scheme, Brownian bridge barrier detection, and control variates. Calibration involves solving a 3-parameter optimization problem (theta, alpha, rho_bar) for fixed beta and eta. The model has 7 parameters total (theta, alpha, beta, rho_bar, eta, theta_plus, theta_minus) with constraints. The SDE system involves 4 correlated Brownian motions. Empirical estimation of eta from risk reversal beta data adds another layer of complexity. Overall more complex than standard Heston but less than full LSV/SLV models.

The paper provides complete mathematical derivations including the SDE system, Riccati ODE solutions, closed-form characteristic function, and hedging arguments. Monte Carlo methodology is described (Andersen 2008 QE scheme, Brownian bridge barrier detection). However, empirical calibration uses proprietary JPMorgan FX options data (EURUSD, USDJPY risk reversals 2022-2025), and no code or synthetic data is provided. Parameter values for examples are given (S=100, r=q=0, T=0.25, theta=0.01, alpha=0.3, beta=2). Reproducing the empirical analysis requires access to the same market data.