betaFeedback

WorksMathematics

Positivity and Boundedness of Solutions to the Guyon-Lekeufack Volatility Model with General Kernels

Abstract

We study nonnegativity of the volatility in the continuous-time Guyon-Lekeufack model, in which volatility is an affine function of the convolution of past returns with a kernel K1K_1 and the square root of the integral of past squared returns weighted by a kernel K2K_2. We show that, when volatility is nonnegative and K1K_1 admits a resolvent of the first kind, uniform boundedness of integrals of K2K_2 with respect to this resolvent yields a deterministic upper bound for the asset price, which is then a bounded martingale on every finite time interval. This result is the counterpart of the strict local martingality criterion for negative volatility in Guyon, Jeannin and Jourdain (2026, Corollary 2.6(ii)). We also derive such bounds for popular stochastic volatility models when the Brownian motions driving the asset price and volatility have correlation −1-1. We give sufficient conditions for nonnegative volatility with a completely monotone return convolution kernel K1K_1, and necessary conditions when K1K_1 and K2K_2 are respectively MM- and NN-exponential kernels. For M=1M=1, we obtain a necessary and sufficient condition.

Acknowledgements: The authors thank Paul Invernizzi for insightful discussions. To sharpen an earlier sufficient condition for nonnegative volatility, the authors proposed using the resolvent of the first kind and a semimartingale decomposition. They used OpenAI’s GPT-6 Astra model to help adapt the proof to this approach; the sharp bounds for exponential kernels had already been derived by the authors.

Introduction

Empirical evidence suggests that the volatility of financial markets depends on the past evolution of the asset price [25, 26, 10, 8, 6]. Guyon and Lekeufack [17] introduced a path-dependent volatility (PDV) model based on past returns and past squared returns weighted by memory kernels. We study when its volatility1 stays nonnegative. From a modeling perspective, verifying the nonnegativity of volatilities is important because it ensures that the correlations specified between the driving Brownian motions retain their interpretation as instantaneous correlations between asset returns (in multi-asset models) or between asset returns and changes in some extra stochastic processes, such as stochastic interest rates or stochastic variance (in single- or multi-asset models). For a Brownian motion WW and deterministic continuous functions g1:R+→Rg_1:\mathbb{R}_+\to\mathbb{R} and g2:R+→[0,∞)g_2:\mathbb{R}_+\to[0,\infty), we define the PDV model through the coupled system, for t≥0t\geq0,

dStSt=σt dWt(PDV)\frac{dS_t}{S_t}=\sigma_t\,dW_t \quad\text{(PDV)}
σt:=β0+β1R‾1,t+β2R‾2,t,R‾1,t:=g1(t)+∫0tK1(t−s)σs dWs,R‾2,t:=g2(t)+∫0tK2(s,t)σs2 ds,\begin{aligned} \sigma_t &:= \beta_0 + \beta_1\overline{R}_{1,t} + \beta_2\sqrt{\overline{R}_{2,t}},\\ \overline{R}_{1,t} &:= g_1(t) + \int_0^t K_1(t-s)\sigma_s\,\mathrm{d}W_s,\\ \overline{R}_{2,t} &:= g_2(t) + \int_0^t K_2(s,t)\sigma_s^2\,\mathrm{d}s, \end{aligned}

where β0,β2>0\beta_0,\beta_2 > 0 and β1<0\beta_1 < 0. The nonnegative convolution kernel K1:(0,∞)→[0,∞)K_1 : (0,\infty) \to[0,\infty) and the nonnegative kernel K2:Γ‾→R+K_2 : \overline{\Gamma} \to\mathbb{R}_+ weight past returns and past squared returns, respectively, with Γ‾:={(s,t)∈R+2:s≤t}\overline{\Gamma} := \{(s,t) \in\mathbb{R}_+^2 : s \le t\}.

When K1K_1 and K2K_2 are respectively MM- and NN-exponential kernels (defined below), with integers M,N≥1M,N \ge1, the PDV model admits a finite-dimensional Markovian representation. The kernels K1K_1 and K2K_2 of the four-factor PDV model are both 2-exponential kernels. Existence and pathwise uniqueness of strong solutions were established for the four-factor model in [23 Theorem 1.2] and for the PDV model with more general Volterra kernels in [5 Theorem 1]. For K1(t)=λ1,1e−λ1,1tK_1(t) = \lambda_{1,1}e^{-\lambda_{1,1}t} and K2(s,t)=λ2,1e−λ2,1(t−s)K_2(s,t) = \lambda_{2,1}e^{-\lambda_{2,1}(t-s)}, Guyon and Lekeufack [17 Section 4.1.5] observed that λ2,1<2λ1,1\lambda_{2,1} < 2\lambda_{1,1} ensures nonnegativity when σ0≥0\sigma_0 \ge0. Nutz and Riveros Valdevenito [23 Theorem 2.6] obtained a strictly positive pathwise lower bound when σ0>0\sigma_0 > 0. Andrès and Jourdain [5 Theorem 2] extended the positivity result to possibly nonconvolution return kernels K1(s,t)=g(s)eh(t)K_1(s,t) = g(s)e^{h(t)} and general squared-return kernels K2(s,t)K_2(s,t) satisfying ∂tK2(s,t)−2h′(t)K2(s,t)≥0\partial_tK_2(s,t) - 2h'(t)K_2(s,t) \ge0.

Nonnegativity and invariance for stochastic Volterra equations are studied in [3, 4, 2]. Bondi and Pulido [7] establish Feller-type tests for scalar convolution equations with nonsingular kernels. Their tests do not directly apply to (PDV) because σ\sigma depends on both R‾1\overline{R}_1 and R‾2\overline{R}_2, and the initial inputs depend on time. As in [7 Lemma 1], we derive a semimartingale decomposition for σ\sigma and obtain sufficient conditions for positivity by controlling the drift term arising from memory.

Strikingly, for every T>0T > 0, when σ≥0\sigma\ge0 and condition (11) below holds, Theorem 1 below gives a deterministic upper bound on sup⁡t∈[0,T]St\sup_{t\in[0,T]} S_t, despite the Brownian driving noise. This bound depends only on the kernels, coefficients and initial inputs. Hence SS is a bounded martingale on [0,T][0,T]. Beyond martingality, this bound restricts the support of the price distribution, ensures finiteness of all positive moments, and implies zero call prices for strikes above the bound. We also extend this price-bound principle to several stochastic volatility models in the limiting case ρ=−1\rho= -1 of perfect negative correlation between the Brownian motions driving the asset price and volatility. In the PDV model, this result and the strict local martingality criterion in [19 Corollary 2.6 (ii)] yield a dichotomy according to the sign of volatility: for smooth return kernels K1K_1 admitting a resolvent of the first kind, SS is a strict local martingale on [0,T][0,T] whenever σt0<0\sigma_{t_0} < 0 has positive probability for some t0∈[0,T)t_0 \in[0,T).

For a completely monotone return kernel K1K_1 (defined below), Theorem 2 below gives explicit conditions on the kernels and initial state that ensure nonnegative volatility, together with an explicit process that bounds σ\sigma from below. In particular, its specialization to the four-factor PDV model provides sufficient conditions on both the kernel parameters and the initial factors, see the second row of Table 2 below. It thus addresses the question of finding sufficient conditions for nonnegative volatility in that model, left open by Nutz and Riveros Valdevenito [23 Section 3.2].

K1K_1Sufficient conditions for positivity of volatilityNecessary conditions for positivity of volatility
1-exp2λ1,1≥λ2,eff,2\lambda_{1,1} \ge\lambda_{2,\mathrm{eff}},<br>0≤β0λ1,10 \le\beta_0\lambda_{1,1}<br>+β2(λ1,1−λ2(R2,0)2)R‾2,0+\beta_2\left(\lambda_{1,1}-\frac{\lambda_2(\mathbf{R}_{2,0})}{2}\right)\sqrt{\overline{R}_{2,0}}2λ1,1≥λ2,eff,2\lambda_{1,1} \ge\lambda_{2,\mathrm{eff}},<br>0≤β0λ1,10 \le\beta_0\lambda_{1,1}<br>+β2(λ1,1−λ2(R2,0)2)R‾2,0+\beta_2\left(\lambda_{1,1}-\frac{\lambda_2(\mathbf{R}_{2,0})}{2}\right)\sqrt{\overline{R}_{2,0}}
2-exp2λ1,2≥λ2,eff,2\lambda_{1,2} \ge\lambda_{2,\mathrm{eff}},<br>max⁡{0,λ^1η0(λ1,1)}≤β0λ1,2\max\{0,\widehat{\lambda}_1\eta_0(\lambda_{1,1})\} \le\beta_0\lambda_{1,2}<br>+β2(λ1,2−λ2(R2,0)2)R‾2,0,+\beta_2\left(\lambda_{1,2}-\frac{\lambda_2(\mathbf{R}_{2,0})}{2}\right)\sqrt{\overline{R}_{2,0}},<br>max⁡{0,λ^1η0(λ1,1)}≤β0λ1,2\max\{0,\widehat{\lambda}_1\eta_0(\lambda_{1,1})\} \le\beta_0\lambda_{1,2}<br>+β2(λ1,2−λ2,eff2)R‾2,0+\beta_2\left(\lambda_{1,2}-\frac{\lambda_{2,\mathrm{eff}}}{2}\right)\sqrt{\overline{R}_{2,0}}λ1,eff≥λ2,eff2,\lambda_{1,\mathrm{eff}} \ge\frac{\lambda_{2,\mathrm{eff}}}{2},<br>λ^1η0(λ1,1)≤β0λ1,2\widehat{\lambda}_1\eta_0(\lambda_{1,1}) \le\beta_0\lambda_{1,2}<br>+β2(λ1,2−12λ2(R2,0))R‾2,0+\beta_2\left(\lambda_{1,2}-\frac{1}{2}\lambda_2(\mathbf{R}_{2,0})\right)\sqrt{\overline{R}_{2,0}}
MM-exp2λ1,M≥λ2,eff,2\lambda_{1,M} \ge\lambda_{2,\mathrm{eff}},<br>max⁡1≤p≤M∑i=pM−1λ^iη0(λ1,i)≤β0λ1,M\max_{1\le p\le M}\sum_{i=p}^{M-1}\widehat{\lambda}_i\eta_0(\lambda_{1,i}) \le\beta_0\lambda_{1,M}<br>+β2(λ1,M−λ2(R2,0)2)R‾2,0,+\beta_2\left(\lambda_{1,M}-\frac{\lambda_2(\mathbf{R}_{2,0})}{2}\right)\sqrt{\overline{R}_{2,0}},<br>max⁡1≤p≤M∑i=pM−1λ^iη0(λ1,i)≤β0λ1,M\max_{1\le p\le M}\sum_{i=p}^{M-1}\widehat{\lambda}_i\eta_0(\lambda_{1,i}) \le\beta_0\lambda_{1,M}<br>+β2(λ1,M−λ2,eff2)R‾2,0+\beta_2\left(\lambda_{1,M}-\frac{\lambda_{2,\mathrm{eff}}}{2}\right)\sqrt{\overline{R}_{2,0}}λ1,eff≥λ2,eff2,\lambda_{1,\mathrm{eff}} \ge\frac{\lambda_{2,\mathrm{eff}}}{2},<br>∑i=1M−1λ^iη0(λ1,i)≤β0λ1,M\sum_{i=1}^{M-1}\widehat{\lambda}_i\eta_0(\lambda_{1,i}) \le\beta_0\lambda_{1,M}<br>+β2(λ1,M−12λ2(R2,0))R‾2,0+\beta_2\left(\lambda_{1,M}-\frac{1}{2}\lambda_2(\mathbf{R}_{2,0})\right)\sqrt{\overline{R}_{2,0}}

Table 2. Conditions from Proposition 3 and Theorem 3 when K2K_2 is an NN-exponential kernel. The deterministic initial factors and inputs satisfy (5) and R‾2,0=g2(0)>0\overline{R}_{2,0}=g_2(0)>0 as in (1). The sufficient conditions assume σ0≥0\sigma_0\ge0, and the necessary conditions assume σ0>0\sigma_0>0. The sufficient and necessary conditions coincide for M=1M=1 when σ0>0\sigma_0>0. For M≥2M\ge2, the sufficient and necessary rate conditions involve λ1,M\lambda_{1,M} and λ1,eff\lambda_{1,\mathrm{eff}}, respectively; the sufficient conditions on the initial factors involve both λ2,eff\lambda_{2,\mathrm{eff}} and λ2(R2,0)\lambda_2(\mathbf{R}_{2,0}), while the necessary condition involves λ2(R2,0)\lambda_2(\mathbf{R}_{2,0}). For M≥3M\ge3, the sufficient conditions bound the maximum of zero and all tail sums ∑i=pM−1λ^iη0(λ1,i)\sum_{i=p}^{M-1}\widehat{\lambda}_i\eta_0(\lambda_{1,i}), whereas the necessary condition bounds only the full sum corresponding to p=1p=1.

References

  1. [1]E. Abi Jaber, M. Larsson, and S. Pulido. Affine Volterra processes. The Annals of Applied Probability, 29(5):3155–3200, 2019. doi: 10.1214/19-AAP1477. URL https://doi.org/10.1214/19-AAP1477.
  2. [2]E. Abi Jaber, C. Cuchiero, L. Pelizzari, S. Pulido, and S. Svaluto-Ferro. Polynomial Volterra processes. Electronic Journal of Probability, 29(176):1–37, 2024. doi: 10.1214/24-EJP1234. URL https://doi.org/10.1214/24-EJP1234.
  3. [3]A. Alfonsi. Nonnegativity preserving convolution kernels. Application to stochastic Volterra equations in closed convex domains and their approximation. Stochastic Processes and their Applications, 181:104535, 2025. doi: 10.1016/j.spa.2024.104535. URL https://doi.org/10.1016/j.spa.2024.104535.
  4. [4]A. Alfonsi and G. Szulda. On non-negative solutions of stochastic Volterra equations with jumps and non-Lipschitz coefficients. Bernoulli, 31(4):2890–2915, 2025. doi: 10.3150/24-BEJ1830. URL https://doi.org/10.3150/24-BEJ1830.
  5. [5]H. Andrés and B. Jourdain. Existence, uniqueness and positivity of solutions to the Guyon–Lekeufack path-dependent volatility model with general kernels. International Journal of Theoretical and Applied Finance, 28(7-8):2550019, 2025. doi: 10.1142/S0219024925500190. URL https://doi.org/10.1142/S0219024925500190.
  6. [6]H. Andrés, A. Boumezoued, and B. Jourdain. The implied volatility surface (also) is path-dependent. Quantitative Finance, 26(7):1017–1047, 2026. doi: 10.1080/14697688.2026.2637739. URL https://doi.org/10.1080/14697688.2026.2637739.
  7. [7]A. Bondi and S. Pulido. Feller’s test for explosions of stochastic Volterra equations. Electronic Journal of Probability, 31(102):1–41, 2026. doi: 10.1214/26-EJP1561. URL https://doi.org/10.1214/26-EJP1561.
  8. [8]J.-P. Bouchaud. Radical complexity. Entropy, 23(12):1676, 2021. doi: 10.3390/e23121676. URL https://doi.org/10.3390/e23121676.
  9. [9]A. Bourdon and T. Jeannin. Moments in Rough Bergomi and boundary attainment in Rough Heston, 2026. URL https://arxiv.org/abs/2606.07482. arXiv:2606.07482.
  10. [10]R. Chicheportiche and J.-P. Bouchaud. The fine-structure of volatility feedback I: Multi-scale self-reflexivity. Physica A, 410:174–195, 2014. doi: 10.1016/j.physa.2014.05.007. URL https://doi.org/10.1016/j.physa.2014.05.007.
  11. [11]D. Darrow and G. Stepaniants. A spectral theory of scalar Volterra equations. Communications of the American Mathematical Society, 6:634–725, 2026. doi: 10.1090/cams/64. URL https://doi.org/10.1090/cams/64.
  12. [12]O. El Euch and M. Rosenbaum. The characteristic function of rough Heston models. Mathematical Finance, 29(1):3–38, 2019. doi: 10.1111/mafi.12173. URL https://doi.org/10.1111/mafi.12173.
  13. [13]D. Filipović, S. Tappe, and J. Teichmann. Invariant manifolds with boundary for jump-diffusions. Electronic Journal of Probability, 19(51):1–28, 2014. doi: 10.1214/EJP.v19-2882. URL https://doi.org/10.1214/EJP.v19-2882.
  14. [14]M. Friesen, S. Gerhold, and K. Wiedermann. Boundary behaviour of the Volterra square-root process, 2026. URL https://arxiv.org/abs/2606.07290. arXiv:2606.07290.
  15. [15]G. Gazzani and J. Guyon. Pricing and calibration in the α-factor path-dependent volatility model. Quantitative Finance, 25(3):471–489, 2025. doi: 10.1080/14697688.2025.2472892. URL https://doi.org/10.1080/14697688.2025.2472892.
  16. [16]G. Gripenberg, S.-O. Londen, and O. Staffans. Volterra Integral and Functional Equations, volume 34 of Encyclopedia of Mathematics and its Applications. Cambridge University Press, 1990. doi: 10.1017/CBO9780511662805. URL https://doi.org/10.1017/CBO9780511662805.
  17. [17]J. Guyon and J. Lekeufack. Volatility is (mostly) path-dependent. Quantitative Finance, 23(9):1221–1258, 2023. doi: 10.1080/14697688.2023.2221281. URL https://doi.org/10.1080/14697688.2023.2221281.
  18. [18]J. Guyon and L. Parent. Reconciling P- and Q-calibration with path-dependent volatility. Risk, September 2026. URL https://www.risk.net/cutting-edge/7964125/reconciling-p-and-q-calibration-with-path-dependent-volatility.
  19. [19]J. Guyon, T. Jeannin, and B. Jourdain. The martingale property in Volterra volatility models, 2026. URL https://arxiv.org/abs/2610.32684. arXiv:2610.32684.
  20. [20]P. S. Hagan, D. Kumar, A. S. Lesniewski, and D. E. Woodward. Managing smile risk. Wilmott Magazine, pages 84–108, 2002. URL https://lesniewski.us/papers/published/ManagingSmileRisk.pdf.
  21. [21]S. L. Heston. A closed-form solution for options with stochastic volatility with applications to bond and currency options. The Review of Financial Studies, 6(2):327–343, 1993. doi: 10.1093/rfs/6.2.327. URL https://doi.org/10.1093/rfs/6.2.327.
  22. [22]A. Jacquier. Advanced methods in derivatives pricing with application to volatility modelling, 2017. URL https://www.ma.imperial.ac.uk/~ajacquie/IC_AMDP/IC_AMDP_Docs/AMDP.pdf. Lecture notes, Imperial College London.
  23. [23]M. Nutz and A. Riveros Valdevenito. On the Guyon–Lekeufack volatility model. Finance and Stochastics, 28(4):1203–1223, 2024. doi: 10.1007/s00780-024-00544-2. URL https://doi.org/10.1007/s00780-024-00544-2.
  24. [24]R. Schöbel and J. Zhu. Stochastic volatility with an Ornstein–Uhlenbeck process: An extension. European Finance Review, 3(1):23–46, 1999. doi: 10.1023/A:1009803506170. URL https://doi.org/10.1023/A:1009803506170.
  25. [25]G. Zumbach. Time reversal invariance in finance. Quantitative Finance, 9(5):505–515, 2009. doi: 10.1080/14697680802616712. URL https://doi.org/10.1080/14697680802616712.
  26. [26]G. Zumbach. Volatility conditional on price trends. Quantitative Finance, 10(4):431–442, 2010. doi: 10.1080/14697680903266730. URL https://doi.org/10.1080/14697680903266730.

Paper details

vlnc:W_2n6vhzt3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

OpenAI’s GPT-6 Astra model

View statement
Looking for related papers…
Contents