A Rough Instantaneous Squared VIX Model for Term Structure and Option Pricing
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.
Ruth: We're looking at the paper: "A Rough Instantaneous Squared VIX Model for Term Structure and Option Pricing".
Warren: Here's the gist of the paper. The paper proposes a rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into the conventional Markovian framework to derive an explicit term structure and a semi-closed-form pricing formula for VIX options The results support a two-step calibration procedure that estimates memory, mean reversion, and date-specific initial states from VIX term-structure indices before calibrating the remaining distributional parameters to VIX option prices Model Introduction The RISVIX model extends the ISVIX framework of Luo et al.
Ruth: (2019) by introducing a fractional kernel K(t) = t(α-1/Γ(α)) into the instantaneous squared VIX dynamics This leads to an explicit VIX term structure governed by a Mittag–Leffler decay function and a semi-closed pricing formula for VIX options based on an affine Fourier–Laplace transform The model is defined by the process Vt = V0 + Z t 0 n K(t - s) n κ(θs - Vs) ds + σV p Vs dB1,s + dJ y s o, θt = θ0 + σθ Z t 0 p θs dB2,s The parameters satisfy κ, µy > 0 and α ∈ (1/2, 1) Term Structure Derivation The squared VIX is represented by VIX 2 t,τ = τ-1 E Q t
R t+τ t ds: Under the RISVIX model, the terminal squared VIX is given by Equation 11, which shows that the time-zero squared VIX curve moves monotonically from V0 towards θ0 The diffusion and jump parameters do not enter this curve, but they affect the distribution of future VIX through the history term in Equation 11 Option Pricing Formula The option pricing is achieved through the Fourier–Laplace transform f(u; 0, T) = E Q 0 h exp n u VIX 2 T,τ0 o i Theorem 3.2 guarantees that the VIX transform is finite on every vertical contour u = η + iξ with η ∈ (0, ηbarT,τ0) The resulting option value is obtained by solving one nonlinear Volterra equation for each transform argument and evaluating a one-dimensional numerical integral Empirical Validation The empirical study uses daily VIX term-structure observations and VIX option quotes from January 2018 to June 2025 RISVIX improves the fit of the VIX term structure, with larger gains for its slope and curvature and during periods of elevated volatility It also improves out-of-sample VIX option pricing, reducing MSE and MAE by 13.06% and 6.33%, respectively, with larger gains at shorter maturities and in high-VIX states Model Comparison RISVIX achieves the lowest MSE and MAE among the five models in both periods Relative to LZZ, the full-sample MSE reductions are 10.42%, 17.21% and 13.18% for VIX levels, slopes and curvature, respectively The option-pricing results show a similar pattern The paper concludes that RISVIX improves term-structure fit, particularly during periods of market stress Over January–June 2020, these reductions increase to 46.59% for curvature The option-pricing results show a similar pattern With global parameters fixed at their training-sample estimates and date-specific initial states updated from the contemporaneous VIX term structure, RISVIX reduces out-of-sample MSE and MAE by 13.06% and 6.33%, respectively The gains are larger for shorter-maturity options and high-VIX states Future work could extend RISVIX to a joint SPX–VIX framework to impose consistency across equity and volatility markets References Abi Jaber, E., Weak existence and uniqueness for affine stochastic Volterra equations with L 1-kernels Bernoulli, 2021, 27, 1583–1615 Abi Jaber, E., Illand, C.
Warren: and Li, S.X., Joint SPX & VIX Calibration with Gaussian Polynomial Volatility Models: Deep Pricing with Quantization Hints Abi Jaber, E., Larsson, M.
Ruth: and Pulido, S., Affine Volterra Processes Bardgett, C., Gourier, E.
Title and authors: Ruth: So we're looking at this paper now, "A Rough Instantaneous Squared VIX Model for Term Structure and Option Pricing." It seems like the title already tells you what it's about: using a rough model to figure out the term structure of volatility and then pricing options on top of that.
Warren: Yeah, and the authors are Tianfeng Yang, Wen Su, Blanka Horvath. They're taking something that was previously Markovian—that standard Markovian framework—and they're adding fractional memory into it to get this rough instantaneous squared VIX model.
Iris: That sounds like they're trying to account for some roughness in the volatility dynamics that standard models miss, which is a big deal when you’re trying to price options.
Ruth: Exactly. The summary says they derive an explicit representation of the VIX term structure governed by a Mittag-Leffler decay function and get a semi-closed-form pricing formula using an affine Fourier–Laplace transform for VIX options.
Warren: That's what they are promising, right? A way to get this structure without having to solve all that messy differential equations every single time you want to price something.
Iris: And they support a two-step calibration procedure, which means they estimate the memory and mean reversion parameters from VIX term structure indices first, before calibrating the distributional parameters using VIX option prices.
Ruth: That separation is key, because it allows them to focus on the structure of time dependence—the memory—before getting bogged down in estimating all those volatility shocks and jumps.
Warren: And they used daily VIX term structure indices and option quotes from January two thousand eighteen to June two thousand twenty-five for their empirical study, which is a solid look at recent market behavior <ref:2611.07120#pg1,option quotes from January 2018 to June 2025>.
Iris: The results show that this RISVIX model improves the fit of the VIX term structure, especially when volatility is elevated, with larger gains in both its slope and its curvature during those high-volatility periods.
Ruth: That makes sense because they are specifically designed to capture that history dependence, which is what drives things during stress.
Warren: And for option pricing, they say it improves out-of-sample results by reducing the mean squared error by thirteen point zero six percent and the mean absolute error by six point three three percent when compared to the Markovian benchmark in out-of-sample testing, with those gains being bigger at shorter maturities and in high VIX states.
Iris: That reduction in error suggests that this new model is indeed providing a better fit for real option pricing than the standard approach across different market conditions.
Ruth: The paper also compares it to other models, saying RISVIX achieves the lowest mean squared error and mean absolute error among five models tested in both periods.
Warren: They show relative MSE reductions of ten point four two percent for VIX levels, seventeen point two one percent for slopes, and thirteen point one eight percent for curvature when compared to LZZ, which is a Markovian model with alpha equal to one in that comparison.
Title and authors: Iris: When they look specifically at the periods of market stress, like January through June two thousand twenty those improvements increase even more significantly for curvature by about forty-six point five nine percent.
Ruth: That jump in performance during high-volatility states is what really stands out for me from an economist's view—the model seems to actually be capturing the dynamics when things get scary.
Warren: And they’re not just talking about the fit; they’re talking about a practical pricing formula derived through that Fourier–Laplace transform, which means you can price options directly from those term structure estimates.
Iris: The methodology relies on solving a nonlinear Volterra equation for the psi one and a scalar Riccati ordinary differential equation for psi two to get that final pricing formula.
Ruth: So, to put it simply, this paper is proposing a new way to model how volatility changes over time by adding fractional memory so we can better understand the term structure and price derivatives more accurately.
Warren: And the implication there is that you can build a more robust pricing tool for VIX options without needing to solve overly complex equations for every single market condition.
Iris: The paper also points out that they found gains from fractional memory are particularly pronounced when volatility is elevated, reaching about forty point nine three percent reduction in MSE when the VIX level is greater than or equal to twenty-five.
Ruth: That’s a concrete number showing that this isn't just theoretical improvement; it shows real predictive power during those high-volatility events we expect to see sometimes.
Warren: And they are open about the limitations, which is important for me as a trader—they state that they fix the global parameters at their training sample estimates and update the date specific initial states from the contemporaneous VIX term structure.
Iris: They also say that future work could extend RISVIX to a joint SPX–VIX framework, which would help impose consistency across equity and volatility markets, addressing a limitation of modeling VIX in isolation.
Ruth: So, the big picture is that this paper provides a more detailed and empirically validated tool for understanding volatility over time and pricing derivatives based on that structure.
Warren: It gives us a way to build better risk management tools by using these term structure estimates directly to price things, rather than relying on just the spot VIX level.
Iris: We need to keep an eye on how well this performs when we move outside the training period, because the paper is focused on fitting those specific historical data points.
Ruth: Well, that’s where we end for today with "A Rough Instantaneous Squared VIX Model for Term Structure and Option Pricing." It's a solid piece of work trying to bridge the gap between complex memory effects and practical derivative pricing.
The paper's summary: Ruth: So, this paper proposes a rough instantaneous squared VIX model by adding fractional memory to the standard Markovian framework to get an explicit term structure and a semi-closed formula for VIX options.
Warren: Right, so they’re using fractional memory—that t power alpha minus one over gamma—to build this model that shows how the VIX changes over time, which is usually messy.
Iris: They're saying this leads to an explicit VIX term structure governed by a Mittag-Leffler decay function and a semi-closed pricing formula based on an affine Fourier–Laplace transform.
Ruth: That’s the core mechanism here, trying to get that term structure curve without having to solve those complicated equations every time you want a price.
Warren: The summary points out they use a two-step calibration procedure—they estimate memory and mean reversion parameters first using VIX term structure indices before calibrating the distributional parameters using VIX option prices.
Iris: That separation is what makes it interesting for me, because it means we can nail down the time dependence of volatility before we get into estimating all those market shocks and jumps.
Ruth: It sounds like they are trying to build a more robust system where the structure of time itself informs the pricing, rather than just relying on a standard Markovian assumption.
Warren: And empirically, they tested this model using daily VIX term structure observations and option quotes from January two thousand eighteen to June two thousand twenty-five.
Iris: What did the empirical study actually show us about how it fits the data?
Ruth: It improved the fit of the VIX term structure, particularly during periods of elevated volatility, with larger gains in both its slope and its curvature.
Warren: And for option pricing, they reported reducing mean squared error by thirteen point zero six percent and mean absolute error by six point three three percent compared to their benchmark models in out-of-sample testing.
Iris: Those percentage reductions are significant because they show that this approach provides better pricing results, especially for shorter maturities and during high volatility states.
Ruth: So, the paper suggests that fractional memory isn't just a theoretical curve; it has tangible benefits when modeling real market stress and pricing derivatives.
Warren: The model also highlights that the gains from fractional memory are particularly pronounced when volatility is elevated, with out-of-sample mean squared error reductions reaching forty point nine three percent when the VIX level is greater than or equal to twenty five.
Iris: That tells me that this model really shines when things get scary in the market, which is where we need better tools for risk management.
Ruth: It also mentions a limitation—they fixed the global parameters at their training sample estimates and updated date specific initial states from the contemporaneous VIX term structure.
Warren: And they're looking ahead, saying future work could extend RISVIX to a joint SPX and VIX framework to impose consistency across equity and volatility markets.
Iris: That’s a big thing—linking the two markets so we don't have separate models for them, which is something we definitely need to consider.
Ruth: So, this paper gives us a more detailed tool for understanding how volatility evolves over time and pricing options based on that evolution.
The paper's improvements: Tom: So, this paper is about taking that rough instantaneous squared VIX model and showing exactly how it improves things compared to older models like LZZ.
Ruth: It shows that when volatility is high, say above twenty five, the reductions in mean squared error jump significantly to forty point nine three percent for curvature.
Warren: That’s a pretty big number, and it suggests that this fractional memory approach really delivers predictive power precisely when we need it most—during those high-volatility events.
Iris: I noticed they also showed improvements across the board in out-of-sample testing, with mean squared error drops of thirteen point zero six percent for VIX levels and six point three three percent for absolute error.
Ruth: So, it's not just better in one area; it’s a more consistent improvement across different types of errors when you test how well the model predicts things outside its original training set.
Warren: And those gains are bigger for shorter maturities and in high VIX states, which means if you're pricing short-term options during a nervous market, this model is giving you a better shot at getting an accurate price.
Iris: I see the methodology relies on solving a nonlinear Volterra equation for psi one and a scalar Riccati ordinary differential equation for psi two to get that final pricing formula.
Ruth: That’s the technical heavy lifting—it shows that by linking the term structure directly to those equations, you get this tractable semi-closed pricing formula instead of just having to do messy numerical integration every time.
Warren: It's about making the process efficient for a quantitative trader; if you can use a closed form, it means less computational cost when you’re running simulations or pricing derivatives quickly.
Iris: The paper also addresses the parameter separation, where they estimate common term structure parameters like alpha and kappa first, before calibrating those volatility innovations like sigma V and lambda Y.
Ruth: That two-step calibration is what I liked; it makes the estimation of memory and mean reversion parameters much more robust because you’re isolating those structural dynamics from the noise in the volatility shocks.
Warren: So, when everyone runs this model, they have to be careful—they have to fix those global parameters based on training data and update those date specific initial states using the current VIX term structure.
Iris: That limitation is important; it means the model's accuracy depends heavily on how well you estimate those initial states from the contemporaneous term structure.
Ruth: And they flag that for future work, extending this to a joint SPX and VIX framework would be key to imposing consistency across equity and volatility markets.
Warren: That links back to what we talked about before; if we can get that cross-market link, it could lead to more integrated risk management tools for the whole portfolio.
Conclusion: Ruth: So we're wrapping up this look at "A Rough Instantaneous Squared VIX Model for Term Structure and Option Pricing." Basically, the paper presents a way to model volatility dynamics using fractional memory to build a more accurate term structure curve and pricing formula for VIX options.
Warren: It’s essentially using that fractional kernel to get an explicit representation of how the squared VIX changes over time, which is usually hidden in complex stochastic processes.
Iris: And they found that this method leads to a two-step calibration procedure where you estimate memory and mean reversion parameters from market indices before calibrating the distributional parameters using option prices.
Ruth: It’s a lot of work upfront, but it allows for a much more robust estimation process when you are dealing with time-dependent structures in markets.
Warren: The results show that this model improves the fit of the VIX term structure significantly, especially during periods of market stress like January through June two thousand twenty.
Iris: I just want to check the identification here; the variation doing the work seems to be tied to that fractional memory term and how it feeds into that Mittag-Leffler decay function.
Ruth: That’s right; it’s all about isolating what variation is actually driving these improved fits, which is a good thing for understanding volatility incentives.
Warren: And for pricing, they claim a reduction in mean squared error by thirteen point zero six percent and mean absolute error by six point three three percent compared to the Markovian benchmark.
Iris: Those numbers are solid, but remember the caveat they mentioned—the model fixes the global parameters based on training sample estimates and updates initial states from contemporaneous VIX term structure data.
Ruth: Exactly, so we have to be careful about how much we rely on those fixed parameters when moving out of the training period.
Warren: The paper also points out that their gains from fractional memory are especially pronounced when volatility is elevated, reaching forty point nine three percent reduction in mean squared error when the VIX level is twenty five or higher.
Iris: That suggests a strong predictive benefit during high-stress regimes, which means better tools for people managing risk in those scary times.
Ruth: It really shows how incorporating memory effects can translate into better practical pricing and forecasting tools for volatility derivatives.
Warren: So, the implication there is that if you want to price VIX options accurately under stress, moving beyond a simple Markovian assumption toward something with fractional memory might be the way forward.
Iris: We need to keep an eye on their future work regarding that joint SPX and VIX framework they mentioned earlier, because consistency across markets is a big goal.
Ruth: It’s definitely worth watching how they extend this to link equity and volatility markets together for a more complete picture of risk.
Tianfeng Yang, Wen Su, Blanka Horvath
Department of Mathematical Sciences, Tsinghua University · Mathematical Institute, University of Oxford
q-fin.PR, q-fin.MF
Submitted: 2026-11-04
Updated: 2026-11-04
License: http://creativecommons.org/publicdomain/zero/1.0/
Importance score: 91/100
The gist: The paper proposes a rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into the conventional Markovian framework to derive an explicit term structure and a…
Key concepts
- Fractional Memory Kernel K(t)
- This kernel, defined as t^(α-1/Γ(α)), introduces fractional memory into the model. Instead of standard Markovian assumptions, it allows the volatility process to depend on its entire history, capturing complex path-dependent behaviors in market volatility.
- VIX Term Structure
- The model derives an explicit formula for how the squared VIX changes over time. This structure is governed by a Mittag–Leffler decay function, providing a continuous representation of the expected future squared VIX based on its past evolution.
- Affine Fourier–Laplace Transform
- This mathematical tool is used to derive a semi-closed-form pricing formula for VIX options. It transforms the complex stochastic process into a form where the option value can be calculated using numerical integration of a one-dimensional equation.
- Two-Step Calibration Procedure
- The model uses a two-step approach: first, estimate memory and mean reversion parameters from observed VIX term structure indices. Second, calibrate the remaining distributional parameters using actual VIX option prices to achieve accurate pricing.
Terminology
Summary
The paper proposes a rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into the conventional Markovian framework to derive an explicit term structure and a semi-closed-form pricing formula for VIX options The results support a two-step calibration procedure that estimates memory, mean reversion, and date-specific initial states from VIX term-structure indices before calibrating the remaining distributional parameters to VIX option prices
Model Introduction
The RISVIX model extends the ISVIX framework of Luo et al. (2019) by introducing a fractional kernel K(t) = t(α-1/Γ(α)) into the instantaneous squared VIX dynamics This leads to an explicit VIX term structure governed by a Mittag–Leffler decay function and a semi-closed pricing formula for VIX options based on an affine Fourier–Laplace transform The model is defined by the process Vt = V0 + Z t 0 n K(t - s) n κ(θs - Vs) ds + σV p Vs dB1,s + dJ y s o, θt = θ0 + σθ Z t 0 p θs dB2,s The parameters satisfy κ, µy > 0 and α ∈ (1/2, 1)
Term Structure Derivation
The squared VIX is represented by VIX 2 t,τ = τ-1 E Q t [R t+τ t ds] Under the RISVIX model, the terminal squared VIX is given by Equation 11, which shows that the time-zero squared VIX curve moves monotonically from V0 towards θ0 The diffusion and jump parameters do not enter this curve, but they affect the distribution of future VIX through the history term in Equation 11
Option Pricing Formula
The option pricing is achieved through the Fourier–Laplace transform f(u; 0, T) = E Q 0 h exp n u VIX 2 T,τ0 o i Theorem 3.2 guarantees that the VIX transform is finite on every vertical contour u = η + iξ with η ∈ (0, ηbarT,τ0) The resulting option value is obtained by solving one nonlinear Volterra equation for each transform argument and evaluating a one-dimensional numerical integral
Empirical Validation
The empirical study uses daily VIX term-structure observations and VIX option quotes from January 2018 to June 2025 RISVIX improves the fit of the VIX term structure, with larger gains for its slope and curvature and during periods of elevated volatility It also improves out-of-sample VIX option pricing, reducing MSE and MAE by 13.06% and 6.33%, respectively, with larger gains at shorter maturities and in high-VIX states
Model Comparison
RISVIX achieves the lowest MSE and MAE among the five models in both periods Relative to LZZ, the full-sample MSE reductions are 10.42%, 17.21% and 13.18% for VIX levels, slopes and curvature, respectively The option-pricing results show a similar pattern
Conclusion
The paper concludes that RISVIX improves term-structure fit, particularly during periods of market stress Over January–June 2020, these reductions increase to 46.59% for curvature The option-pricing results show a similar pattern With global parameters fixed at their training-sample estimates and date-specific initial states updated from the contemporaneous VIX term structure, RISVIX reduces out-of-sample MSE and MAE by 13. p. 0 These results support a two-step
calibration procedure that estimates memory, mean reversion, and date-specific initial states
from VIX term-structure indices before calibrating the remaining distributional parameters
to VIX option prices. Using daily VIX term-structure indices and VIX option quotes from
January 2018 to June 2025, we find that RISVIX improves upon the Markovian benchmark
in fitting the VIX term structure, with larger gains during elevated-volatility periods and
for its slope and curvature.Where the paper says it
References
Abi Jaber, E., Weak existence and uniqueness for affine stochastic Volterra equations with L 1-kernels Bernoulli, 2021, 27, 1583–1615 Abi Jaber, E. p. 5 2 gives E
Q[
R T
0
(V
2
t + θ
2
t
) dt] < ∞ and establishes that
θ is a square-integrable martingale. General weak-existence results for affine stochastic Volterra
equations can be found in Abi Jaber (2021). Existing models for VIX derivatives can be organized into two broad approaches. The first
derives VIX from the joint risk-neutral dynamics of the S&P 500 and its latent variance process,
thereby maintaining a structural link between SPX and VIX derivatives, but often requiring a
relatively rich specification and joint calibration to both markets (Zhang and Zhu 2006, Zhu and
Zhang 2007, Cont and Kokholm 2013, Lin et al. The remainder of the paper is organized as follows. Section 2 reviews the ISVIX framework,
fractional Brownian motion and affine Volterra processes. Existing models for VIX derivatives can be organized into two broad approaches. The first
derives VIX from the joint risk-neutral dynamics of the S&P 500 and its latent variance process,
thereby maintaining a structural link between SPX and VIX derivatives, but often requiring a
relatively rich specification and joint calibration to both markets (Zhang and Zhu 2006, Zhu and
Zhang 2007, Cont and Kokholm 2013, Lin et al. Tradable exposure is instead provided by VIX futures and
options, which allow investors to hedge volatility risk and take directional positions on changes
in market uncertainty (Luo et al. At the same time, evidence that volatility exhibits
rough, non-Markovian dynamics has prompted growing interest in the implications of roughness
for VIX futures and options (Gatheral et al. A Rough Instantaneous Squared VIX Model
for Term Structure and Option Pricing
Tianfeng Yang†
, Wen Su∗‡
, Blanka Horvath‡
†Department of Mathematical Sciences, Tsinghua University, China
‡Mathematical Institute, University of Oxford, OX2 6GG, UK
Abstract
The VIX term structure provides forward-looking measures of market uncertainty across
horizons, while VIX options allow investors to hedge volatility risk. This paper proposes a
rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into
the conventional Markovian framework. The remainder of the paper is organized as follows. Section 2 reviews the ISVIX framework,
fractional Brownian motion and affine Volterra processes. The
closest study is Ye et al. (2026), who derive the VIX term structure from rough Heston using a
Markovian approximation. Existing models for VIX derivatives can be organized into two broad approaches. The first
derives VIX from the joint risk-neutral dynamics of the S&P 500 and its latent variance process,
thereby maintaining a structural link between SPX and VIX derivatives, but often requiring a
relatively rich specification and joint calibration to both markets (Zhang and Zhu 2006, Zhu and
Zhang 2007, Cont and Kokholm 2013, Lin et al. Existing models for VIX derivatives can be organized into two broad approaches. The first
derives VIX from the joint risk-neutral dynamics of the S&P 500 and its latent variance process,
thereby maintaining a structural link between SPX and VIX derivatives, but often requiring a
relatively rich specification and joint calibration to both markets (Zhang and Zhu 2006, Zhu and
Zhang 2007, Cont and Kokholm 2013, Lin et al. Comparisons with additional benchmarks from the recent
literature also show improvements to varying degrees in both VIX term-structure fitting and
option pricing. Keywords: VIX; VIX term structure; VIX options; Affine Volterra model
JEL Classification: G12, G13
1 Introduction
Since the global financial crisis of 2008, the VIX has become a central gauge of financial-market
uncertainty and the reference index for a major market in volatility derivatives (Whaley 2000,
Luo et al. Under this re-indexing, the
relevant local model interval for an option with time to expiration T is [0, T + τ0], corresponding
to the calendar interval [t0, t0 + T + τ0]. Such re-indexing is standard in rough-volatility pricing;
depending on the model, conditioning at the new origin is encoded through an initial forward
curve or another model-specific initial input (El Euch and Rosenbaum 2018, Horvath et al. 2016, Gatheral et al. 2018, El Euch and Rosenbaum 2019). 2016, Gatheral et al. 2018, El Euch and Rosenbaum 2019). 2016, Gatheral et al. 2018, El Euch and Rosenbaum 2019). The
closest study is Ye et al. (2026), who derive the VIX term structure from rough Heston using a
Markovian approximation. 2019, Eraker and Yang
2022). The second approach models variance, volatility, VIX, or VIX futures directly (Lin 2007,
Dupoyet et al. A Rough Instantaneous Squared VIX Model
for Term Structure and Option Pricing
Tianfeng Yang†
, Wen Su∗‡
, Blanka Horvath‡
†Department of Mathematical Sciences, Tsinghua University, China
‡Mathematical Institute, University of Oxford, OX2 6GG, UK
Abstract
The VIX term structure provides forward-looking measures of market uncertainty across
horizons, while VIX options allow investors to hedge volatility risk. This paper proposes a
rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into
the conventional Markovian framework. (3)
It has stationary increments and admits sample paths that are locally Hölder continuous of every
order strictly less than H. The case H = 1/2 is standard Brownian motion, whereas H < 1/2
gives rougher paths. A Rough Instantaneous Squared VIX Model
for Term Structure and Option Pricing
Tianfeng Yang†
, Wen Su∗‡
, Blanka Horvath‡
†Department of Mathematical Sciences, Tsinghua University, China
‡Mathematical Institute, University of Oxford, OX2 6GG, UK
Abstract
The VIX term structure provides forward-looking measures of market uncertainty across
horizons, while VIX options allow investors to hedge volatility risk. This paper proposes a
rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into
the conventional Markovian framework. A Rough Instantaneous Squared VIX Model
for Term Structure and Option Pricing
Tianfeng Yang†
, Wen Su∗‡
, Blanka Horvath‡
†Department of Mathematical Sciences, Tsinghua University, China
‡Mathematical Institute, University of Oxford, OX2 6GG, UK
Abstract
The VIX term structure provides forward-looking measures of market uncertainty across
horizons, while VIX options allow investors to hedge volatility risk. This paper proposes a
rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into
the conventional Markovian framework. At the same time, evidence that volatility exhibits
rough, non-Markovian dynamics has prompted growing interest in the implications of roughness
for VIX futures and options (Gatheral et al. 2018, Jacquier et al. The second approach models variance, volatility, VIX, or VIX futures directly (Lin 2007,
Dupoyet et al. 2011, Luo and Zhang 2012, Mencía and Sentana 2013, Huskaj and Nossman 2013,
Luo et al. The second approach models variance, volatility, VIX, or VIX futures directly (Lin 2007,
Dupoyet et al. 2011, Luo and Zhang 2012, Mencía and Sentana 2013, Huskaj and Nossman 2013,
Luo et al. (2011) develop a direct volatility-futures model and compare CEV and CIR specifications
with and without jumps. Huskaj and Nossman (2013) model the term structure of VIX futures
directly rather than deriving it from an equity-variance process, and Mencía and Sentana (2013)
conduct an extensive empirical evaluation of VIX derivative valuation models across different
market conditions. At the same time, evidence that volatility exhibits
rough, non-Markovian dynamics has prompted growing interest in the implications of roughness
for VIX futures and options (Gatheral et al. 2018, Jacquier et al. A Rough Instantaneous Squared VIX Model
for Term Structure and Option Pricing
Tianfeng Yang†
, Wen Su∗‡
, Blanka Horvath‡
†Department of Mathematical Sciences, Tsinghua University, China
‡Mathematical Institute, University of Oxford, OX2 6GG, UK
Abstract
The VIX term structure provides forward-looking measures of market uncertainty across
horizons, while VIX options allow investors to hedge volatility risk. This paper proposes a
rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into
the conventional Markovian framework. Building on this representation, Luo et al. (2019)
develop a unified Markovian ISVIX model with a stochastic mean level and jumps for pricing
VIX derivatives across horizons. First, it contributes to
the literature on direct modeling of VIX term structures and derivatives. Lin (2007) price VIX
futures by integrating information under the physical and risk-neutral measures, while Dupoyet
et al. Keywords: VIX; VIX term structure; VIX options; Affine Volterra model
JEL Classification: G12, G13
1 Introduction
Since the global financial crisis of 2008, the VIX has become a central gauge of financial-market
uncertainty and the reference index for a major market in volatility derivatives (Whaley 2000,
Luo et al. The index itself is not directly tradable: it is calculated from S&P 500 option
prices and summarizes the market’s risk-neutral expectation of S&P 500 variation over the next
30 days (Luo and Zhang 2012). Keywords: VIX; VIX term structure; VIX options; Affine Volterra model
JEL Classification: G12, G13
1 Introduction
Since the global financial crisis of 2008, the VIX has become a central gauge of financial-market
uncertainty and the reference index for a major market in volatility derivatives (Whaley 2000,
Luo et al. The index itself is not directly tradable: it is calculated from S&P 500 option
prices and summarizes the market’s risk-neutral expectation of S&P 500 variation over the next
30 days (Luo and Zhang 2012). Comparisons with additional benchmarks from the recent
literature also show improvements to varying degrees in both VIX term-structure fitting and
option pricing. Keywords: VIX; VIX term structure; VIX options; Affine Volterra model
JEL Classification: G12, G13
1 Introduction
Since the global financial crisis of 2008, the VIX has become a central gauge of financial-market
uncertainty and the reference index for a major market in volatility derivatives (Whaley 2000,
Luo et al. Keywords: VIX; VIX term structure; VIX options; Affine Volterra model
JEL Classification: G12, G13
1 Introduction
Since the global financial crisis of 2008, the VIX has become a central gauge of financial-market
uncertainty and the reference index for a major market in volatility derivatives (Whaley 2000,
Luo et al. The index itself is not directly tradable: it is calculated from S&P 500 option
prices and summarizes the market’s risk-neutral expectation of S&P 500 variation over the next
30 days (Luo and Zhang 2012). Keywords: VIX; VIX term structure; VIX options; Affine Volterra model
JEL Classification: G12, G13
1 Introduction
Since the global financial crisis of 2008, the VIX has become a central gauge of financial-market
uncertainty and the reference index for a major market in volatility derivatives (Whaley 2000,
Luo et al. The index itself is not directly tradable: it is calculated from S&P 500 option
prices and summarizes the market’s risk-neutral expectation of S&P 500 variation over the next
30 days (Luo and Zhang 2012). A Rough Instantaneous Squared VIX Model
for Term Structure and Option Pricing
Tianfeng Yang†
, Wen Su∗‡
, Blanka Horvath‡
†Department of Mathematical Sciences, Tsinghua University, China
‡Mathematical Institute, University of Oxford, OX2 6GG, UK
Abstract
The VIX term structure provides forward-looking measures of market uncertainty across
horizons, while VIX options allow investors to hedge volatility risk. This paper proposes a
rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into
the conventional Markovian framework. 2 on [0, T]. Then, the following identity holds almost surely:
E
Q
0
[Vs] = θ0 + (V0 − θ0)Eα,1(−κsα
), (10)
where Ea,b(z) = P∞
n=0 z
n/Γ(an+b) is the two-parameter Mittag–Leffler function. ISVIX coincides with
the instantaneous diffusive variance in the absence of price jumps and otherwise includes a jump
correction (Luo et al. 2 Fractional Brownian Motion and Fractional Kernels
A standard fractional Brownian motion (fBM) with Hurst parameter H ∈ (0, 1) is a centred
Gaussian process BH with covariance (Mandelbrot and Van Ness 1968)
E
Q
h
B
H
s B
H
t
i
=
1
2
s
2H + t
2H − |t − s|
2H
, s, t ≥ 0. Under the risk-neutral measure Q, the
RISVIX model is
Vt = V0 +
Z t
0
K(t − s)
n
κ(θs − Vs) ds + σV
p
Vs dB1,s + dJ
y
s
o
,
θt = θ0 + σθ
Z t
0
p
θs dB2,s, 0 ≤ t ≤ T. (9)
Here V0, θ0 ≥ 0 are F0-measurable initial states satisfying E
Q[V
2
0 + θ
2
0
] < ∞, and B1 and B2 are
independent Brownian motions. The second approach models variance, volatility, VIX, or VIX futures directly (Lin 2007,
Dupoyet et al. 2011, Luo and Zhang 2012, Mencía and Sentana 2013, Huskaj and Nossman 2013,
Luo et al. 2018, Rømer 2022). The
prominence of the VIX market, together with this evidence, creates a need for models that can
capture the dynamics of the VIX while remaining sufficiently tractable for derivative pricing. Existing models for VIX derivatives can be organized into two broad approaches. The first
derives VIX from the joint risk-neutral dynamics of the S&P 500 and its latent variance process,
thereby maintaining a structural link between SPX and VIX derivatives, but often requiring a
relatively rich specification and joint calibration to both markets (Zhang and Zhu 2006, Zhu and
Zhang 2007, Cont and Kokholm 2013, Lin et al. Comparisons with additional benchmarks from the recent
literature also show improvements to varying degrees in both VIX term-structure fitting and
option pricing. Keywords: VIX; VIX term structure; VIX options; Affine Volterra model
JEL Classification: G12, G13
1 Introduction
Since the global financial crisis of 2008, the VIX has become a central gauge of financial-market
uncertainty and the reference index for a major market in volatility derivatives (Whaley 2000,
Luo et al. The
closest study is Ye et al. (2026), who derive the VIX term structure from rough Heston using a
Markovian approximation. Existing models for VIX derivatives can be organized into two broad approaches. The first
derives VIX from the joint risk-neutral dynamics of the S&P 500 and its latent variance process,
thereby maintaining a structural link between SPX and VIX derivatives, but often requiring a
relatively rich specification and joint calibration to both markets (Zhang and Zhu 2006, Zhu and
Zhang 2007, Cont and Kokholm 2013, Lin et al. Existing models for VIX derivatives can be organized into two broad approaches. The first
derives VIX from the joint risk-neutral dynamics of the S&P 500 and its latent variance process,
thereby maintaining a structural link between SPX and VIX derivatives, but often requiring a
relatively rich specification and joint calibration to both markets (Zhang and Zhu 2006, Zhu and
Zhang 2007, Cont and Kokholm 2013, Lin et al.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Technical Lemmas
For κ > 0, the resolvent Rκ of κK satisfies Rκ + κK ∗ Rκ = κK For a locally square-integrable martingale M, write (K ∗ dM)t = R t 0 K(t − s) dMs Lemma A.3 states that the weight w in Equation 11 is strictly decreasing on (0, ∞), with w(0+) = 1 and w(∞) = 0 For u ∈ C, consider the complex Riccati equation ψ'2(s; u) = κψ 1(s; u) + σ squared θ squared ψ 2(s; u) squared, ψ 2(0; u) = u[1 - w(τ0)]
Proofs
The proof of Proposition 3.1 shows that the conditional mean of ISVIX is given by Equation 10 The proof of Theorem 3. p. 6 By Theorem 3. 1, its terminal squared VIX is given by Equation 11 evaluated at
t = T and τ = τ0.Where the paper says it
Corollary Proof
The proof for Corollary 3.1 shows that E Q 0 h (VIX T,τ0 − K) + i = 1/4 sqrt(π/2z 3/2) erfc(K sqrt(z)) f(η + iξ; 0, T) dξ = 1/2 sqrt(π/2z 3/2) Re "erfcK sqrt η + iξ (η + iξ) 3/2 f(η + iξ; 0, T) dξ The second equality follows from conjugate symmetry
Appendix A.3
The proof of the bound for psi 1 and psi 2 shows that ψ 1(s; u) squared ≤ u squared aτ0(0) squared + h squared T α Γ(1 + α) Eα,1 [1 + σ squared V pηinfinity] T α The proof of the bound for psi 2 shows that ψ 2(s; u) <= 2η h 1 - w(τ0) + 2κT aτ0L∞ The final step proves that E Q 0 [e c X T] < ∞ whenever 2eCT c < 1
Appendix B
The proof of the martingale property for real u = η ∈ [0, ηbarT,τ0) shows that E is a uniformly integrable martingale on [0, T] The exponential formula for the marked Poisson process shows that D is a strictly positive martingale with E Q 0 [D 2 T] = exp (λy Z T 0 [ϱ(2pη(s)) − 2ϱ(pη(s)) + 1] ds) The Bayes’ formula therefore yields E Q 0 [ET] = 1, so E is a uniformly integrable martingale on [0, T]
Appendix C
The proof of Corollary 3.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Fractional memory integration enables rough volatility modeling that improves VIX term structure fit, particularly during periods of market stress, by introducing a fractional Volterra kernel directly into the ISVIX framework of Luo et al. (2019). This allows the model to capture history dependence and low local regularity observed in volatility dynamics without relying on a Markovian approximation.
-
The introduction of an explicit VIX term structure formula governed by Mittag-Leffler decay, derived from Equation 10, provides a unique continuous representative of the conditional mean for the instantaneous squared VIX process. This allows for a two-step calibration procedure where memory and mean reversion parameters are estimated from VIX indices before distributional parameters are calibrated to option prices.
-
The affine Fourier-Laplace transform formulation yields a semi-closed-form pricing formula for VIX options, reducing Mean Squared Error (MSE) by 13.06% and Mean Absolute Error (MAE) by 6.33% compared to the Markovian benchmark in out-of-sample testing. This provides a tractable and efficient method for pricing VIX derivatives directly from the term structure estimates.
-
The model supports a parameter separation where common term-structure parameters (alpha, kappa) and date-specific initial states (Vt, θt) are estimated first from VIX indices before calibrating volatility innovations (sigmaV, sigmaθ, lambdaY, muY). This two-step calibration strategy makes the estimation of memory and mean reversion parameters more robust across different market regimes.
-
The model demonstrates that gains from fractional memory are particularly pronounced when volatility is elevated, with out-of-sample MSE reductions reaching 40.93% when VIXt is greater than or equal to 25. This indicates that the model offers significant predictive power during high-volatility events compared to benchmark models like LZZ (alpha = 1).
-
The model can be extended to a joint SPX–VIX framework, allowing for consistency across equity and volatility markets, which is a crucial future step for developing more comprehensive risk management tools. This extension addresses the limitation of modeling VIX in isolation by imposing structural links between the two markets.
Abstract
The VIX term structure provides forward-looking measures of market uncertainty across horizons, while VIX options allow investors to hedge volatility risk. This paper proposes a rough instantaneous squared VIX (RISVIX) model by introducing fractional memory into the conventional Markovian framework. We derive an explicit representation of the VIX term structure governed by Mittag--Leffler decay and a semi-closed-form pricing formula for VIX options based on an affine Fourier--Laplace transform. These results support a two-step calibration procedure that estimates memory, mean reversion, and date-specific initial states from VIX term-structure indices before calibrating the remaining distributional parameters to VIX option prices. Using daily VIX term-structure indices and VIX option quotes from January 2018 to June 2025, we find that RISVIX improves upon the Markovian benchmark in fitting the VIX term structure, with larger gains during elevated-volatility periods and for its slope and curvature. RISVIX also improves out-of-sample option pricing, reducing MSE and MAE by 13.06% and 6.33%, respectively, with larger gains at shorter maturities and in high-VIX states. These findings suggest that the gains from fractional memory are particularly pronounced when volatility is elevated and can be obtained within a tractable direct model of VIX derivatives. Comparisons with additional benchmarks from the recent literature also show improvements to varying degrees in both VIX term-structure fitting and option pricing.
Related papers
- Multiple Stopping and Exercise Boundaries for American Put Options
- (Early) AI Compute Asset Pricing
- Reaction-boundary variance and adjoint-consistent local-volatility projection
- A Semi-Analytic Approach To Valuing Auto-Callable Accrual Notes
- Valuation Model of Chinese Convertible Bonds Based on Monte Carlo Simulation
- Information-Theoretic Model for the Growth Optimal Portfolio of Stocks