Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Semi-Static Variance-Optimal Hedging of Covariance Risk in Multi-Asset Derivatives

Konstantinos Chatziandreou, Sven Karbach

Abstract

We develop a semi-static framework for the variance-optimal hedging of multi-asset derivatives exposed to correlation and covariance risk. The approach combines continuous-time dynamic trading in the underlying assets with a static portfolio of auxiliary contingent claims. Using a multivariate Galtchouk--Kunita--Watanabe decomposition, we show that the resulting global mean-variance problem decouples naturally into an inner continuous-time projection onto the space spanned by the underlying assets and an outer finite-dimensional quadratic optimization over the static hedging instruments. To systematically select suitable auxiliary claims, we leverage multidimensional functional spanning theory, establishing that otherwise unhedgeable cross-gamma exposures can be structurally mitigated through static strips of vanilla, product, and spread options. As a central application, we derive explicit semi-static replication formulas for covariance swaps and geometric dispersion trades. Our framework accommodates a broad class of asset dynamics, including quadratic and stochastic Volterra covariance models, as well as affine stochastic covariance models with jumps, yielding tractable semi-closed-form solutions via Fourier transform techniques. Extensive numerical experiments demonstrate that incorporating optimally weighted static strips of cross-asset instruments substantially reduces the mean-squared hedging error relative to purely dynamic benchmark strategies across various model classes.

Semi-Static Variance-Optimal Hedging of Covariance Risk in Multi-Asset Derivatives

Abstract

We develop a semi-static framework for the variance-optimal hedging of multi-asset derivatives exposed to correlation and covariance risk. The approach combines continuous-time dynamic trading in the underlying assets with a static portfolio of auxiliary contingent claims. Using a multivariate Galtchouk--Kunita--Watanabe decomposition, we show that the resulting global mean-variance problem decouples naturally into an inner continuous-time projection onto the space spanned by the underlying assets and an outer finite-dimensional quadratic optimization over the static hedging instruments. To systematically select suitable auxiliary claims, we leverage multidimensional functional spanning theory, establishing that otherwise unhedgeable cross-gamma exposures can be structurally mitigated through static strips of vanilla, product, and spread options. As a central application, we derive explicit semi-static replication formulas for covariance swaps and geometric dispersion trades. Our framework accommodates a broad class of asset dynamics, including quadratic and stochastic Volterra covariance models, as well as affine stochastic covariance models with jumps, yielding tractable semi-closed-form solutions via Fourier transform techniques. Extensive numerical experiments demonstrate that incorporating optimally weighted static strips of cross-asset instruments substantially reduces the mean-squared hedging error relative to purely dynamic benchmark strategies across various model classes.

Paper Structure

This paper contains 47 sections, 6 theorems, 300 equations, 16 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Literature and Methodological Context
  3. Summary of Main Contributions
  4. Organization of the Paper
  5. Variance-Optimal Pricing and Hedging of Multi-Asset Contingent Claims
  6. Market Setup and Model Dynamics
  7. Class 1: Affine Stochastic Covariance Models with Jumps.
  8. Class 2: Quadratic Covariance Models.
  9. Class 3: Stochastic Volterra Covariance Models.
  10. Introducing leverage.
  11. The Semi-Static Variance-Optimal Hedging Problem
  12. Galtchouk-Kunita-Watanabe Decomposition
  13. Solution to the Semi-Static Hedging Problem
  14. Multidimensional Spanning and the Static Replication of Covariance Risk
  15. Covariance Swaps
  16. ...and 32 more sections

Key Result

Proposition 3.1

Let $h:\mathbb{R}_+^n\to\mathbb{R}$ and fix a reference point $\bm a=(a_1,\dots,a_n)\in\mathbb{R}_+^n$. Assume that $h$ admits continuous partial derivatives $\partial^\alpha h$ for all multi-indices $\alpha=(\alpha_1,\dots,\alpha_n)$ in and satisfies appropriate growth and decay conditions at infinity such that the required infinite-strike integrals are well-defined (see Cui04052022 for explicit

Figures (16)

  • Figure 1: Greedy-forward sparse variance-optimal frontiers for the covariance-swap hedge. The figure reports the progressive improvement (decrease) of the mean-squared hedging error as the cardinality $m$ of the static portfolio increases. The steep initial improvement followed by a pronounced plateau is consistent with the quadratic structure \ref{['eq:quadratic-form']}: a small number of auxiliary instruments captures most of the projection of $L_T^0$ onto $\mathrm{span}(L_T^1,\dots,L_T^n)$, while additional instruments are largely redundant because of strong collinearity across nearby strikes and related payoff families.
  • Figure 2: Stepwise composition of the sparse portfolio in the vanilla + log-spread family. Rows correspond to the active cardinality $m$, and columns correspond to option strikes. The figure shows that the optimizer combines a small number of stable vanilla strikes on the two marginals with a highly concentrated set of log-spread options near the central spread region. This is consistent with the decomposition implied by \ref{['eq:polarizationquadraticvariation']}, where the covariance direction is represented as a signed correction to marginal variance directions.
  • Figure 3: Stepwise composition of the sparse portfolio in the vanilla + geometric-spread family. The structure closely mirrors the log-spread case, but now the auxiliary correction is expressed in the ratio coordinate $Z=S_1/S_2$. Numerically, this family behaves as a discrete proxy for the spread-variance correction, implemented in a multiplicative rather than additive coordinate system.
  • Figure 4: Payoff decomposition of the full vanilla + log-spread basket. The left panel reports the marginal vanilla contribution as a function of the diagonal level $S_1=S_2=S$, while the right panel reports the spread contribution as a function of $Z=\log(S_1/S_2)$. The unconstrained spread component is centered and sign-sensitive, which matches its role as the corrective covariance direction in \ref{['eq:polarizationquadraticvariation']}; the long-only spread component is nearly suppressed, illustrating the loss of the necessary signed hedge.
  • Figure 5: Payoff decomposition of the full vanilla + geometric-spread basket. As in the log-spread case, the marginal vanilla block captures the dominant convexity along the diagonal, while the ratio-option block provides the covariance-specific correction in the relative coordinate $Z=S_1/S_2$. The unconstrained portfolio therefore preserves a centered relative-move hedge, whereas the long-only portfolio loses most of this correction.
  • ...and 11 more figures

Theorems & Definitions (19)

  • Proposition 3.1: Multidimensional Spanning, cf. Cui04052022
  • Example 3.2: Geometric Covariance Swap Replication via Product Options
  • Proposition 3.3: Inverse-square kernels from local vega-flatness
  • proof
  • Example 3.4: Geometric Covariance Swap Replication via Log-Spread Options
  • Remark 3.5: Instrument Selection and Liquidity Constraints
  • Remark 3.6: On the choice of kernel in variance and covariance swaps
  • Proposition 4.1: Fourier pricing identity
  • Theorem 4.2: Fourier Representation of the Multivariate GKW Decomposition
  • proof
  • ...and 9 more