Pricing Options on Forwards in Function-Valued Affine Stochastic Volatility Models
2025
Read Pricing Options on Forwards in Function-Valued Affine Stochastic Volatility Models (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-14813.txt.
Introduction
Jian He, Sven Karbach, and Asma Khedher price options on forward contracts in models where whole price curves move at once, in a 2025 paper. In energy and interest-rate markets, traders deal in forward curves, not single prices. The authors treat the curve as an infinite-dimensional object driven by a stochastic volatility process.
They analyze two model families. One is Gaussian, steered by a matrix-valued noise process. The other adds jumps whose size depends on the current state, extending a classic framework. For both, they find conditions guaranteeing that the needed averages exist, then derive semi-closed pricing formulas for plain call and put options using Fourier methods. The point is practicality: even with infinitely many risk factors, prices tied to specific maturities and the curve's shape can still be computed. That matters for markets where term structure risk is the whole game.
Catalog record
| YEAR | 2025 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 107 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-14813
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Pricing Options on Forwards in Function-Valued Affine Stochastic Volatility Models Thanks: Asma Khedher is grateful for the financial support by the Research Foundation Flanders (FWO) under the grant FWO WOG W001021N Abstract . We study the pricing of European-style options written on forward contracts within function-valued infinite-dimensional affine stochastic volatility models. The dynamics of the underlying forward price curves are modeled within the Heath-Jarrow-Morton-Musiela framework as solution to a stochastic partial differential equation modulated by a stochastic volatility process. We analyze two classes of affine stochastic volatility models: (i) a Gaussian model governed by a finite-rank Wishart process, and (ii) a pure-jump affine model extending the Barndorff–-Nielsen–Shephard framework with state-dependent jumps in the covariance component. For both models, we derive conditions for the existence of exponential moments and develop semi-closed Fourier-based pricing formulas for vanilla call and put options written on forward price curves. Our approach allows for tractable pricing in models with infinitely many risk factors, thereby capturing maturity-specific and term structure risk essential in forward markets. Keywords: option pricing, stochastic volatility models, affine processes, forward curve dynamics, commodity derivatives, Wishart processes, Fourier methods, HJM framework, exponential moments. Jian He Sven Karbach Asma Khedher [ [email protected] ] [ [email protected] ] [ [email protected] ] 1. Introduction We consider a tractable framework for modeling forward price and forward rate curves in commodity and fixed-income markets, and develop semi-closed pricing formulas for the valuation of European vanilla options written on (potentially) the entire forward curve. Following the Heath–Jarrow–Morton–Musiela (HJMM) paradigm, originally proposed for interest-rate markets by [ 29 ] and later transferred to commodity markets by [ 11 , 3 ] , we describe forward prices directly in time-to-maturity coordinates. In this representation, the forward curve evolves according to a hyperbolic stochastic partial differential equation (SPDE) driven by a cylindrical Brownian motion whose covariance is itself modeled by a stochastic process that we call the instantaneous covariance process . We investigate two affine specifications of the covariance process: (i) An extension of the finite-rank Wishart process to Hilbert–Schmidt operators introduced in [ 14 ] and (ii) a state-dependent jump extension of the operator-valued Ornstein–Uhlenbeck process introduced in [ 16 ] . As both models are from the affine class, their associated Fourier-Laplace transforms admit semi-explicit expressions in terms of the solutions to a set of generalized Riccati equations , which allows for semi-closed pricing formulas for vanilla call and put options. The derivation of these pricing formulas, together with rigorous conditions for the existence of exponential moments and a numerical analysis, constitut the principal contribution of this article. 1.1. The HJMM framework For any time t\geq 0 and maturity date T\geq t , the forward price F(t,T) represents the agreed-upon price at time t for a transaction occurring at the future date T . Utilizing the Musiela parametrization x\equiv T-t , forward prices become functions of time-to-maturity. We define the logarithmic forward prices as f_{t}(x)\equiv\ln(F(t,t+x)) , where x,t\geq 0 , and call the mapping x\mapsto f_{t}(x) the (logarithmic) forward price curve at time t\geq 0 . The forward curve dynamics are described by the stochastic partial differential equation (SPDE): (1.1) \displaystyle\begin{cases}\mathop{}\!\textnormal{d}f_{t}(x)&=\left(\frac{\partial}{\partial x}f_{t}(x)+g_{t}(x)\right)\mathop{}\!\textnormal{d}t+\sum_{i=1}^{d}\sigma_{t}^{(i)}(x)\mathop{}\!\textnormal{d}W_{t}^{(i)},\\ f_{0}(x)&=F(0,x),\end{cases} where (W^{(i)}_{t})_{0\leq t\leq T} , {i=1},\ldots,{d} , are independent Brownian motions. The drift term g_{t}(x) satisfies a HJM no-arbitrage condition (see [ 9 , Sec. 2.4] and Section 3 below), ensuring that each forward price process (F(t,T))_{0\leq t\leq T} is a martingale under some risk-neutral measure \mathbb{Q} . The functions \{\sigma_{t}^{(i)}(x))_{0\leq t\leq T} , {i=1},\ldots,{d} , are referred to as the instantaneous volatilities of the forward curve. They characterize the local covariance structure between forward prices of different maturities. The integer d\in\mathbb{N}\cup\{+\infty\} is called the rank of the noise and corresponds to the number of independent risk factors driving the forward curves. If d<\infty , we refer to ( 1.1 ) as a finite-rank HJM model; otherwise, it is an infinite-rank HJM model. 1.2. Stochastic covariance models for forward curve dynamics The diffusion term in ( 1.1 ) can be interpreted in a more abstract form as the stochastic integral of an operator-valued process (\sigma_{t})_{0\leq t\leq T} acting on the increments of a cylindrical Brownian motion (\mathrm{d}W_{t})_{0\leq t\leq T} . We understand (\sigma_{t})_{0\leq t\leq T} as the stochastic volatility operator, encoding the time-varying covariance structure of the forward curve dynamics. The functions \sigma_{t}^{(i)} can therefore be interpreted as the principal components of \sigma_{t} , capturing the volatility structure along different directions in the underlying space of forward curves. The joint evolution of the forward curve and its stochastic volatility, represented by the process (f_{t},\sigma_{t})_{0\leq t\leq T} , thus defines an infinite-dimensional stochastic volatility model for forward curve dynamics. The motivation for adopting a stochastic volatility framework stems from empirical evidence in commodity markets: volatility is time-varying, exhibits clustering, and is prone to abrupt spikes; see, e.g., [ 24 , 6 ] . Classical models with constant volatility cannot capture these phenomena, whereas stochastic volatility models allow for a more realistic and flexible description of market behavior. Furthermore, to accurately capture volatility smiles in commodity option markets, incorporating stochastic volatility is essential. In contrast to approaches that perform a priori dimension reduction of the covariance, we do not restrict the rank of the volatility operator based on principal component analysis (PCA) or other variance-based criteria. Instead, we adopt a flexible, full-rank modeling approach, retaining all potential sources of risk. This is motivated by the observation that discarding components solely on the basis of explained variance can result in suboptimal choices for higher-order, non-linear functionals, such as those relevant for pricing, hedging, or risk management [ 12 ] . Note that even when the true rank is only moderately high, our methodology first emphasizes maximal expressiveness and generality. Finite-rank approximations are introduced a posteriori, and we study convergence of option prices as the rank increases. Empirical studies support this design choice: for instance, [ 1 ] and [ 34 ] demonstrate that power forward markets exhibit substantial idiosyncratic risk, and that high-dimensional noise is needed to capture the observed variability. In particular, the latter shows that even in PCA, around 10 factors are typically required to explain the majority of the variance, underlining the importance of accommodating high-dimensional dynamics from the outset. Here our function-valued formulation helps, as it is essentially non-parametric, relying on the full initial forward curve and the full covariance operator as model inputs. The covariance structure is encoded via a positive kernel operator, which can capture empirical features with a minimal set of hyperparameters and enables direct calibration to market data, including option prices and historical term structure dynamics. The stochastic volatility models considered in this paper belong to the class of affine processes , which play a central role in modern financial modeling due to their analytical tractability. Affine processes satisfy the so-called affine transform formula , which expresses the cumulant generating function as an affine function of the initial state and the solution of a system of generalized Riccati ordinary differential equations (ODEs). A key advantage of this framework is the quasi-explicit availability of the Fourier–Laplace transform, which enables efficient and accurate pricing of options and other derivatives via Fourier inversion methods; see [ 10 , 22 , 23 ] . The extension of affine