Martingale property and moment explosions in signature volatility models
Eduardo Abi Jaber, Paul Gassiat, Dimitri Sotnikov · 2025
Read Martingale property and moment explosions in signature volatility models by Eduardo Abi Jaber, Paul Gassiat, Dimitri Sotnikov (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-2503-17103.txt.
Introduction
Eduardo Abi Jaber, Paul Gassiat, and Dimitri Sotnikov study in 2025 when a signature volatility model gives a true martingale price process. In their model, volatility is a linear form of the signature of a time-extended Brownian motion. Martingale pricing is essential for fair asset valuation in finance.
Apart from trivial cases, the price is a true martingale exactly when the linear form has odd order and a correlation parameter is negative. The proof analyzes explosion times of a signature differential equation. Once martingality holds, they also characterize when higher moments exist. Practitioners must choose odd order forms, they warn.
Catalog record
| CREATOR | Eduardo Abi Jaber, Paul Gassiat, Dimitri Sotnikov |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 123 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-17103
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Martingale property and moment explosions in signature volatility models Eduardo Abi Jaber, Paul Gassiat, Dimitri Sotnikov Abstract We study the martingale property and moment explosions of a signature volatility model, where the volatility process of the log-price is given by a linear form of the signature of a time-extended Brownian motion. Excluding trivial cases, we demonstrate that the price process is a true martingale if and only if the order of the linear form is odd and a correlation parameter is negative. The proof involves a fine analysis of the explosion time of a signature stochastic differential equation. This result is of key practical relevance, as it highlights that, when used for approximation purposes, the linear combination of signature elements must be taken of odd order to preserve the martingale property. Once martingality is established, we also characterize the existence of higher moments of the price process in terms of a condition on a correlation parameter. 1 Introduction The martingale property and moment explosions of the asset price process are fundamental considerations in stochastic volatility modeling, particularly due to their implications for derivative pricing and numerical methods. The martingale property is a key criterion for assessing the validity of a stochastic volatility model, as emphasized by Sepp and Rakhmonov (2023) . In particular, strict local martingale models exhibit structural deficiencies, such as the failure of put-call parity and deviations in the asymptotic behavior of implied volatility wings ( Cox and Hobson, 2005 ; Heston, Loewenstein, and Willard, 2007 ; Jacquier and Keller-Ressel, 2018 ) . Similarly, the existence of moments for the price process plays a crucial role in both theoretical and computational aspects of option pricing. In particular, the existence of finite moments is crucial for ensuring the numerical stability and convergence of Monte Carlo methods used to compute derivative prices and their sensitivities. Moments also play a key role in the proof of asymptotic formulae, allowing one to transition from large deviations of the price process to call price asymptotics, see for instance ( Friz et al., 2021 , Section 4.2) and Lee (2004) . Both of these problems have been extensively studied in the literature, dating back to the works of Girsanov (1960) and Novikov (1973) . However, Novikov’s criterion is often not applicable to stochastic volatility models. An alternative approach consists in using a change-of-measure argument to transform the problem into the study of SDE explosion; see, for instance, McKean (1969, Chapter 3.7) . This technique was applied to stochastic volatility modeling by Sin (1998) , who observed that in certain models—such as the Hull–White model—the stock price can be a strict martingale for certain parameter values. In this context, Lions and Musiela (2007) present general sufficient conditions for martingality and strict local martingality when the volatility process follows a stochastic differential equation (SDE), and also address the problem of moment existence. Subsequently, Mijatović and Urusov (2012) provide a more general criterion for martingality in SDE-based models, while Bernard, Cui, and McLeish (2017) extend this work to the general spot-volatility correlation case -1\leq\rho\leq 1 , offering a complete classification of convergence properties for general one-factor Markovian stochastic volatility models. Most of these works, however, focus on Markovian models where both the asset price and volatility processes are solutions of SDEs. For example, Jourdain (2004) proposes a martingality criterion, along with integrability conditions for prices in SABR-like models. Similarly, Andersen and Piterbarg (2007) address the moment problem and provide sufficient martingality conditions for SABR-like and Heston-like models. For the non-Markovian setting, we refer to the works of Criens (2020) and Criens and Glau (2018) , which establish general results on the martingale property of stochastic exponential martingales for Itô processes. However, deriving such results for specific path-dependent models often requires careful, case-by-case analysis. A notable class of such models considers volatility processes governed by stochastic Volterra equations. For the rough Bergomi model, Gassiat (2018) establishes conditions for both martingality and non-martingality, as well as a sufficient condition for moment explosions. Gerhold, Pachschwöll, and Ruf (2024) propose an alternative approach to verifying higher-order moment explosions by examining the integrability of the price supremum and apply this method to the rough Bergomi model. In the context of the Volterra Heston model, Abi Jaber, Larsson, and Pulido (2019) establish conditions for the martingale property, while Gerhold, Gerstenecker, and Pinter (2018) study the problem of moment explosions. In the class of so-called path-dependent volatility (PDV) models ( Guyon and Lekeufack, 2023 ) , Nutz and Riveros Valdevenito (2024) analyze the martingale property in a two-factor PDV model. Additionally, Ruf (2015) provides a general martingality criterion for stochastic exponentials defined as non-anticipative path-dependent functionals of solutions to stochastic differential equations. Despite its generality, this result cannot be directly applied to stochastic volatility models and requires a thorough analysis to verify the condition for each particular model. Recently, signature-based models have gained popularity in non-Markovian modeling. Such models were first studied by Arribas, Salvi, and Szpruch (2020) and further developed by Cuchiero, Gazzani, and Svaluto-Ferro (2023) ; Cuchiero, Gazzani, Möller, and Svaluto-Ferro (2025) , and Abi Jaber and Gérard (2025) . Their distinguishing feature is that the volatility process is represented as a (possibly infinite) linear combination of signature elements. These models can serve both for the approximation, see Cuchiero et al. (2023) , and the exaxt representation, see Abi Jaber and Gérard (2025) , of a wide range of models, though in both cases, the linear combination of signature elements must be truncated for practical use. As we will demonstrate, the truncation order plays a crucial role in determining whether the price process is a true martingale. However, no results on either martingality or moment explosions have been established for these models so far. The aim of this paper is to fill this gap by providing necessary and sufficient conditions for the martingality of the price process and by studying the finiteness of its moments in signature volatility models. We adopt the multiplicative framework of Abi Jaber and Gérard (2025) ; Cuchiero et al. (2025) , in which the price process is given by a stochastic exponential, making the question of martingality non-trivial. The Markovian setting provides key insights into the martingality problem. In particular, for truncated exponential Ornstein-Uhlenbeck models with negative spot-volatility correlation, obtained by truncating the power series expansion of the exponential, the price process is a martingale if and only if the truncation order is odd. A similar pattern appears in polynomial volatility models, where volatility is expressed as a polynomial function of a Brownian motion or an Ornstein–Uhlenbeck process, as in the Quintic model of Abi Jaber, Illand, and Li (2023) . These results can be established using techniques similar to those in Lions and Musiela (2007, Theorem 2.4(i)) . Since path signatures play an analogous role to polynomials on path space, we expect similar characterizations to hold beyond the Markovian framework. In this work, we show that the order of the linear combination of signature elements plays the same role as the order of a polynomial and that a similar result holds for signature volatility models. Specifically, given two Brownian motions, B and W , correlated with coefficient \rho\in[-1,1] , we consider the following signature volatility model