[Paper Review] Introduction into "Local Correlation Modelling"

Alex Langnau|ArXiv.org|Sep 18, 2009

Stochastic processes and financial applications4 references3 citations

TL;DR

This paper introduces Local Correlation Modelling (LCM) as a numerically efficient extension of Dupire's local volatility model to multi-asset derivatives, enabling consistent pricing with both individual and index option markets by making correlations dynamic. The key contribution is identifying 'chewing-gum' moves—correlation shifts that preserve basket and individual distributions but materially affect worst-of options—highlighting their sensitivity to higher-order correlation dynamics beyond the first principal component.

ABSTRACT

In this paper we provide evidence that financial option markets for equity indices give rise to non-trivial dependency structures between its constituents. Thus, if the individual constituent distributions of an equity index are inferred from the single-stock option markets and combined via a Gaussian copula, for example, one fails to explain the steepness of the observed volatility skew of the index. Intuitively, index option prices are encoding higher correlations in cases where the option is particularly sensitive to stress scenarios of the market. As a result, more complex dependency structures emerge than the ones described by Gaussian copulas or (state-independent) linear correlation structures. In this paper we "decode" the index option market and extract this correlation information in order to extend the multi-asset version of Dupire's "local volatility" model by making correlations a dynamic variable of the market. A "local correlation" model (LCM) is introduced for the pricing of multi-asset derivatives. We show how consistency with the index volatility data can be achieved by construction. LCM achieves consistency with both the constituent- and index option markets by construction while preserving the efficiency and easy implementation of Dupire's model.

Motivation & Objective

• To address inconsistencies between index option markets and individual constituent options when using deterministic correlation matrices or Gaussian copulas.
• To model state-dependent correlation dynamics that better reflect market-observed skew and volatility patterns.
• To quantify residual correlation risk in multi-asset derivatives, particularly for worst-of options, that is not captured by basket volatility skew alone.
• To demonstrate that worst-of options are sensitive to higher-order correlation dynamics (e.g., 'chewing-gum' moves), making them valuable for inferring joint asset dynamics.
• To extend Dupire’s local volatility framework to the multi-asset setting with dynamic correlations while preserving calibration to observed option prices.

Proposed method

• Constructs a family of correlation matrices $\hat{\rho}_{ij} \in F(\rho, \xi)$ centered around a base correlation $\rho$, with perturbations controlled by a state variable $u$.
• Uses an analytical inversion of the correlation mapping via Eq. LABEL:ustar to compute the state variable $u$ from observed correlation shifts.
• Pre-computes a lookup table of correlation matrices and their Cholesky decompositions for efficient Monte Carlo simulation.
• Applies the Cholesky decomposition to generate correlated Brownian motions in pathwise simulations, ensuring consistency with the target correlation structure.
• Employs a local volatility framework where each asset’s volatility depends on its spot price, extending Dupire’s single-asset model to multi-asset settings.
• Calibrates the model to both individual option prices and index option prices, ensuring consistency across the entire term structure.

Experimental results

Research questions

• RQ1 How can a multi-asset local volatility model be constructed to be consistent with both individual and index option markets?
• RQ2 What is the impact of correlation dynamics beyond the first principal component on the pricing of exotic multi-asset derivatives?
• RQ3 Why do worst-of options exhibit significant price sensitivity to 'chewing-gum' moves that preserve basket and individual distributions?
• RQ4 To what extent can worst-of options help complete the market and reveal hidden correlation dynamics not observable through index volatility skew?
• RQ5 How can the non-uniqueness of correlation matrix solutions be managed while preserving calibration to observed option prices?

Key findings

• Worst-of put prices differ significantly between two calibration setups: 0.306 (Set 1, centered at market-like correlation) vs. 0.463 (Set 2, identity center) at a 60% strike, despite identical individual and basket distributions.
• The average correlation remains roughly the same across both setups, indicating that basket-level consistency does not imply equivalence in exotic option pricing.
• Worst-of options are highly sensitive to higher-order correlation dynamics (e.g., 'chewing-gum' moves), which preserve the basket and individual distributions but alter the joint dependence structure.
• Only about half of the index volatility skew can be attributed to individual component skews, suggesting that state-dependent correlation dynamics play a crucial role in explaining index behavior.
• The LCM model successfully captures these residual correlation risks by dynamically adjusting correlations based on local market states, enabling more accurate pricing of exotic derivatives.
• The model demonstrates that index options alone are insufficient for inferring the full joint dynamics of a basket, and worst-of options provide critical additional information for market completion.