Volatility models in practice: Rough, Path-dependent or

Markovian?
Eduardo Abi Jaber∗1 and Shaun (Xiaoyuan) Li†2,3
1
Ecole Polytechnique, CMAP
2
AXA Investment Managers
3
Université Paris 1 Panthéon-Sorbonne, CES
arXiv:2401.03345v1 [q-fin.MF] 7 Jan 2024

January 9, 2024

Abstract

An extensive empirical study of the class of Volterra Bergomi models using SPX options
data between 2011 and 2022 reveals the following fact-check on two fundamental claims echoed
in the rough volatility literature:
Do rough volatility models with Hurst index H ∈ (0, 1/2) really capture well
SPX implied volatility surface with very few parameters? No, rough volatility models
are inconsistent with the global shape of SPX smiles. They suffer from severe structural
limitations imposed by the roughness component, with the Hurst parameter H ∈ (0, 1/2)
controlling the smile in a poor way. In particular, the SPX at-the-money skew is incompatible
with the power-law shape generated by rough volatility models. The skew of rough volatility
models increases too fast on the short end, and decays too slow on the longer end where
“negative” H is sometimes needed.
Do rough volatility models really outperform consistently their classical Marko-
vian counterparts? No, for short maturities they underperform their one-factor Markovian
counterpart with the same number of parameters. For longer maturities, they do not sys-
tematically outperform the one-factor model and significantly underperform when compared
to an under-parameterized two-factor Markovian model with only one additional calibratable
parameter.
On the positive side: our study identifies a (non-rough) path-dependent Bergomi model
and an under-parameterized two-factor Markovian Bergomi model that consistently outper-
form their rough counterpart in capturing SPX smiles between one week and three years with
only 3 to 4 calibratable parameters.

JEL Classification: G13, C63, G10, C45.

Keywords: SPX options, Stochastic volatility, Pricing, Calibration, Neural Networks

∗ [email protected]. The first author is grateful for the financial support from the Chaires

FiME-FDD, Financial Risks, Deep Finance & Statistics and Machine Learning and systematic methods in finance
at École Polytechnique.
† [email protected]. The second author is grateful for the financial support provided by AXA Investment

Managers and would like to thank Camille Illand for fruitful discussions and insightful comments.

1
1 Introduction

1.1 Context

In the realm of (rough) volatility modeling, certain claims have gained widespread popularity and
acceptance within academic circles and finance community. Specifically, it has been widely dis-
seminated that rough volatility models exhibit exceptional performance, seemingly reproducing
the stylized facts of option prices with remarkable precision while utilizing only a limited number
of parameters. Furthermore, it has been argued that they outperform their traditional stochastic
volatility model counterparts in capturing the essential features of volatility surfaces. These asser-
tions have been made time and again in various research papers and were often presented with a
high degree of confidence in contrast with little empirical evidence.

There is undoubtedly mathematical beauty in rough volatility1 , and “roughness is in the eye of
the beholder” certainly sounds poetic. However as mathematicians, it is incumbent upon us to
undertake a rigorous examination of these claims before echoing them. First, the assertions about
the superior performance of rough volatility models appear to rest heavily on a limited set of
visual fits confined to specific dates or time intervals, casting doubt on their applicability across
a wider period. Furthermore, there has been notable omission in the literature concerning a
fair and comprehensive comparison between rough volatility models and other models, such as
the conventional Markovian stochastic volatility models and non-rough path-dependent volatility
models. This absence leaves a critical gap in our understanding of the practical usefulness of rough
volatility models, as the non-semimartingale and non-Markovian nature of rough volatility models
have an important implementation cost that needs to be justified.

Recently, a series of independent empirical studies [7, 22, 38, 50], each focusing on different aspects
of the volatility surface, presented evidence against the claim of “superior performance” attributed
to rough volatility models in capturing the main characteristics of volatility surfaces. In particular,
the following observations were made:

• The SPX At-The-Money (ATM) Skew does not follow a power law as prescribed in rough
volatility literature [22, 38];
• Rough volatility models underperform, by a clear margin, their one-factor Markovian coun-
terparts in all market conditions for short maturities with the same number of calibratable
parameters for the joint SPX-VIX calibration problem [7];
• Rough volatility models suffer from severe structural limitations [7, 50].

Inspired by these studies, we conducted our own comprehensive empirical study using daily SPX
implied volatility surface data from CBOE spanning from 2011 to 2022. Our study integrates
various aspects explored in these earlier works and introduces new evidence and arguments to
further challenge the perceived superiority of rough volatility models. We examine both the static
performance of various models, such as their daily fit to the whole SPX volatility surface and to
the ATM skew, as well as their dynamical performance by assessing how well they predict the
future implied volatility surface. Moreover, our empirical study is divided into two parts: part one
is dedicated to options with short maturities (one week to three months), with part two focusing
on options with long maturities (one week to three years). This division offers additional insights
into model performances at different timescales which could be useful in practice.

The empirical experiments in this paper are designed to ensure the comparison between various
models is as fair as possible. To this end, we selected models from the same family of Volterra
1 Fromthe theoretical side, there is no doubt that the rough volatility paradigm has been an inspiring source
of motivation that lead to a better understanding of universal mathematical structures for a large class of non-
Markovian and non-semimartingale Volterra processes, to cite a few [3, 4, 5, 6, 11, 15, 21, 29, 34, 40, 41, 43, 44, 51].

2
Bergomi models where the volatility process is driven by rough, path-dependent (non-rough semi-
martingale) and Markovian factors. The models we ended up with have (almost) the same number
of calibratable parameters that can be interpreted in a similar way. Furthermore, the same daily
forward variance curves extracted from market data are shared among all models for consistency.

As closed-form expression for option pricing under the family of Volterra Bergomi models are
unavailable, we calibrate all models to market data using the generic unified numerical method
‘deep pricing with quantization hints’ developed in [7] based on functional quantization as a first
proxy with Neural Networks (NN) as a corrector. Compared to other neural network structures
used for pricing and calibration [42, 50], this method features significantly lower input dimensions
(i.e. the forward variance curve of the Bergomi model is not part of the NN input). In addition,
it is free from butterfly arbitrage by construction and mesh-free, allowing one to price derivatives
for any strike and maturity combination without interpolation or extrapolation. Notably, this
method also allows us to calibrate each model to the market ATM skew without relying on the
approximation formula of the Bergomi-Guyon expansion [14] as in [22, 38]. This leads to a more
accurate computation of the skew that is free from higher order error in the Bergomi-Guyon
expansion when volatility of volatility (vol of vol) is large.

1.2 What are rough volatility models?

Rough volatility models are a specific subclass of non-Markovian stochastic volatility models, where
the volatility process is assumed to be a non-semimartingale process characterized by continuous
paths rougher than those of standard Brownian motion. Specifically, these models employ vari-
ants of fractional Brownian motion with a Hurst index less than 1/2 to model the instantaneous
volatility. For instance, in [10], the authors used a Riemann-Louiville fractional Brownian motion
given by: Z t
Xt = (t − s)H−1/2 dWs , (1.1)
0
where W is a standard Brownian motion to describe the spot volatility process. The Hurst index
H ∈ (0, 1/2) governs the local regularity of the path of X, indicating that X is Hölder continuous
of order strictly less than H. In the special case of H = 1/2, the process X reduces to classical
Brownian motion. The restriction H > 0 ensures that the convolution kernel t 7→ tH−1/2 is locally
square-integrable so that the stochastic integral is well-defined as an Itô integral.

The Riemann-Liouville fractional Brownian motion (1.1) with H ∈ (0, 1/2) is non-Markovian,
as it represents a weighted average of the increments of Brownian motion by the kernel t 7→
tH−1/2 . It is non-semimartingale due to the singularity of the kernel at t = 0, since H < 1/2.
While historical volatility exhibits many path-dependent features such as volatility clustering,
feedback and memory effects, and multiple timescales [18, 23, 39, 45, 47, 48], the raison d’être
of the rough volatility paradigm is to posit that high fluctuations and erratic/spiky
behavior observed in volatility on short timescales arise from a rough continuous
non-semimartingale process.

To support the necessity of rough volatility and non-semimartingality, two empirical arguments
were presented. First, the logarithm of realized volatility time series for various equity market
indices exhibits statistical “rougher” trajectories (lower Hölder regularity) compared to standard
Brownian motion [33]. An earlier but less popular empirical study of the roughness of the realized
volatility appeared in [20] to motivate the need of a rough multi-fractional Brownian motion.
Second, authors in [10, 17, 27, 31, 33] argued that market ATM skew term structure exhibits
power law decay that explodes in the short term, a phenomenon that rough volatility models can
reproduce [9, 10, 30].

However, the validity of both empirical arguments hinges on exceedingly fine timescales that are
impractical to attain in finite data sets. Regarding the statistical roughness observed in the realized
volatility time series, two independent studies [1, 49] as early as 2018 demonstrated that fast mean-

3
reverting continuous semi-martingales can produce spurious roughness at any realistic timescale,
see also some more recent studies [19, 39] and Section 5.2 below. Recently, challenges to the claimed
explosiveness of the short end of the ATM skew have been raised in [22, 38].

Setting aside the debate, one of our goals is to assess whether the inherent rough non-semimartingale
component in rough volatility models, i.e. the explosiveness of the kernel t 7→ tH−1/2 at t = 0,
aligns convincingly with market volatility surfaces across short and longer maturities as claimed
in the rough volatility literature. It is crucial to emphasize that questioning a rough volatility
model involves challenging its non-semimartingality, i.e. the explosion of the kernel at 0, rather
than disputing its non-Markovian path-dependent nature. Said differently: are rough volatility
models the ‘good’ non-Markovian models in practice?

1.3 Do rough volatility models really outperform significantly their clas-

sical Markovian counterparts?

This paper looks into two main aspects of model performance. First is the static aspect, measured
by how accurate a model fits to the market smiles and ATM skew. Second is the dynamical aspect,
measured by how well a model predicts future implied volatility.

Fitting short maturities between one week to three months: our empirical study shows
that the rough Bergomi model actually underperforms on average compared to the one-factor
Bergomi (Markovian) model with the same number of model parameters in fitting the volatility
surface and the ATM skew. The one-factor Bergomi is also slightly better in predicting future
volatility surface. In other words, there is clearly no advantage in using rough volatility models
over classical Markovian models in the short term.

Fitting longer maturities between one week to three years: the rough Bergomi and the
one-factor Bergomi models produce equally poor fits and are inconsistent with the SPX smiles,
although the one-factor Bergomi model seems to at least be able to achieve a curved shape for
the ATM skew in log-log plot more representative of the market, versus the straight line of rough
volatility models. The rough Bergomi model slightly, but not consistently, outperforms the one-
factor (Markovian) Bergomi model in fitting the volatility surface on average, but scores the highest
variance. For the period between 2017 to 2019, the rough Bergomi model underperforms the one-
factor Bergomi model. Furthermore, the rough Bergomi model significantly underperforms across
all aspects compared to an under-parameterized two-factor (Markovian) Bergomi model with just
one-extra calibratable parameter.

1.4 Do rough volatility models with Hurst index H ∈ (0, 1/2) really cap-
ture SPX smiles with very few parameters?

Our study shows that rough volatility models are inconsistent with the global SPX volatility surface,
caused by their severe structural limitations. In particular, the SPX ATM skew term structure
cannot be modeled by the rigid power-law shape generated by rough volatility models. When one
zooms in on the short end of the volatility surface, the SPX ATM skew does not increase as fast as
H−1/2
√ law decay ∼ T
that of rough volatility models, in line with [22, 38]. In addition, the power for
long maturities observed on the market appears to be faster than 1/ T on average. Specifically,
the average speed of decay of the ATM skew for large maturities over the last decade implies a
negative value of H which is impossible for rough volatility models, since their H is constrained
above zero.

On the other hand, we show that the path-dependent (non-rough) and the under-parameterized
two-factor Bergomi model have no problem capturing the entire term structure of the SPX smiles
and ATM skew via unconstrained H ∈ (−∞, 1/2) and can effectively decouple the short end

4
and the long end of the volatility surface in a parsimonious way, with only 3 to 4 calibratable
parameters.

The superior performance of the non-rough path-dependent Bergomi model versus the rough
Bergomi model highlights that the problem with rough volatility models seems to come from the
non-semimartingale part, i.e. singularity of the kernel at 0. Indeed, the non-rough path-dependent
model is obtained by smoothing out the singularity of the fractional kernel at 0 using a fixed value
ε, which shows that the added flexibility of the path-dependent model comes from eliminating the
rough, non-semimartingale behavior of rough volatility models and not from the calibration of an
additional parameter. In volatility modelling, path-dependency is a desirable trait allowing a model
to be more flexible at capturing several stylized facts of volatility, such as volatility clustering, feed-
back and memory effects, and multiple timescales. However, the non-semimartingality encoded in
rough volatility models via the explosive kernel near t = 0 seems too restrictive and erodes the
flexibility gained from path-dependency. Thus contrary to what is put forward by rough volatil-
ity advocates, roughness/non-semimartingality is not a necessary condition for effective volatility
modelling. Furthermore, all models in this paper can generate spurious ‘statistical roughness’,
see Section 5.2 which suggests that after all, estimation of the roughness of the realized volatility
time series might not be sufficient to quantify the erratic and spiky behavior that characterizes the
volatility of the market.

Outline of the paper: In Section 2, we will first introduce different Volterra Bergomi models
considered in this paper and present their main relevant characteristics. In Section 3, we define the
performance metrics used to evaluate model performance and address the specifics associated with
model calibration. In Section 4, we present the empirical results on the static model performance
in two parts, with part one dedicated to short maturities between one week to three months and
part two for longer maturities between one week to three years. In the final Section 5, we compare
the dynamical aspect of each model by looking at their prediction performance and estimating the
statistical roughness of their simulated volatility paths.

2 The class of Volterra Bergomi models

The general form of the Volterra Bergomi stochastic volatility model for a (forward) stock price S
with instantaneous variance V is defined as:
p
dSt = St Vt dBt ,
1 t 2
 Z 
Vt = ξ0 (t) exp Xt − K (s)ds ,
2 0 (2.1)
Z t
Xt = K(t − s)dWs ,
0
p
with B = ρW + 1 − ρ2 W ⊥ and ρ ∈ [−1, 1]. Here (W, W ⊥ ) is a two-dimensional Brownian motion
on a risk-neutral filtered probability space (Ω, F, (Ft )t≥0 , Q). The process X is a centered Gaussian
process with a non-negative locally square integrable kernel K ∈ L2 ([0, T ], R+ ), in particular
Rt
Xt ∼ N (0, 0 K 2 (u)du), for all t ≤ T . The deterministic input curve ξ0 ∈ L2 ([0, T ], R+ ) allows the
model to match certain term structure of volatility (e.g. the forward variance curve):
Z t  Z t
E Vs ds = ξ0 (s)ds, t ≥ 0.
0 0

The choice of the kernel K defines the dynamics of X (e.g. rough, path-dependent or Markovian)
and plays a crucial role in the model’s capability of capturing the SPX (or other equity indices)
volatility surface. In this paper, We shall consider the following kernels K and their corresponding
model name in Table 1:

5
 Model name K(t) Domain of H Semi-mart. Markovian
rough ηtH−1/2 (0, 1/2] ✗ ✗
path-dependent η(t + ε)H−1/2 (−∞, 1/2] ✓ ✗
H−1/2 −(1/2−H)ε−1 t
one-factor ηε e (−∞, 1/2] ✓ ✓
H−1/2 −(1/2−H)ε−1 t
ηε e +
two-factor −1 (−∞, 1/2] ✓ ✓
ηℓ εHℓ −1/2 e−(1/2−Hℓ )ε t

Table 1: Different kernels K in this paper and their associated Bergomi model name. Note the
two-factor Bergomi model means that the instantaneous variance process V is Markovian in two
dimension driven by a single Brownian motion, see Section 2.4 for detailed explanations.

To ensure comparability among all models, we fix ε = 1/52 and Hℓ = 0.45. This means that
all models contain the same number of calibratable parameters (η, ρ, H), except the two-factor
Bergomi model which takes on an additional calibratable parameter ηℓ . The fixed values may
seem arbitrary, however our goal is not about finding the optimal values for (ε, Hℓ ), but rather to
compare all models as fairly as possible. Furthermore, fixing (ε, Hℓ ) to these values does not alter
the conclusion of this paper compared to setting these parameters free.

The particular parametrization of the one-factor and two-factor Bergomi models ensures that the
parameter H has a similar interpretation across all these models. This would become more clear
as we introduce each Bergomi model in the sections below.

We now introduce the ATM skew which is an important feature of the SPX volatility surface. The
ATM skew ST is defined as:
dbσ (T, k)
ST = ,
dk k=0

where σ(T, k) is the implied volatility of vanilla options calculated by inverting the Black-Scholes
formula with log-moneyness k = log(K/S0 ) and maturity T . For the purpose of this paper, when
we refer to the ATM skew, we would be actually referring to the negative value of ST .

2.1 The rough Bergomi

The rough Bergomi model of [10] defines X as:

Z t
Xt = η (t − s)H−1/2 dWs ,
0

with η > 0 the vol-of-vol parameter and H ∈ (0, 1/2] the Hurst index that coincides with the
roughness of the process X, i.e. the paths of X are Hölder-continuous of any order strictly less
than H, almost surely. For H < 1/2, the fractional kernel K(t) = tH−1/2 explodes as t − → 0,
so that the process X is not a semi-martingale with trajectories rougher than that of standard
Brownian motion. The restriction H > 0 ensures that the kernel K is locally square-integrable, so
that the stochastic convolution is well-defined as an Itô integral.

The rough Bergomi model produces the following ATM skew (assuming flat ξ0 ):

ST ≈ ρηT H−1/2 (2.2)

at the first order of η, see [9, 10, 13, 30]. In particular, the skew explodes at T → 0, with H
controls both the skew explosive rate for small maturities and the rate of power-law decay for large
maturities.

6
2.2 The path-dependent Bergomi

In the path-dependent Bergomi Model, the dynamics of X is governed by:

Z t
Xt = η (t + ε − s)H−1/2 dWs . (2.3)
0

The time shifted kernel, K(t) = (t + ε)H−1/2 , with ε > 0 has been independently introduced over
the years in [2, 7, 13, 37]. It represents a small perturbation in the fractional kernel by ε > 0.
However, this shift means K(0) is finite, thus allowing the domain of H ∈ (−∞, 1/2] to be extended
to −∞ while K remains L2 ([0, T ]) integrable. The process X is now a semi-martingale and thus
the sample paths have the same regularity as standard Brownian motion and are Hölder-continuous
of any order strictly less than 1/2.

The process X is not Markovian. To see this, we apply Itô’s formula and get:
Z t 
dXt = η (t + ε − s)H−3/2 dWs dt + ηεH−1/2 dWt .
0

Next, with the

R t help of the resolvent of the first kind of K, defined as the deterministic measure
L such that 0 K(t − s)L(ds) = 1, for all t ≥ 0, with L(dt) = δ0 (dt)/K(0) + ℓ(t)dt, where ℓ is a
completely monotone function, we get
 Z t 
−1
dXt = − (1/2 − H)ε Xt + f (t − s)Xs ds dt + ηεH−1/2 dWt , (2.4)
0

with
t
εH−3/2
Z
1
f (t) = ℓ(t) + (t + ε − u)H−5/2 ℓ(u)du,
H − 1/2 (H − 1/2)(H − 3/2) 0
Rt
see [2, Lemma 1.2] for detailed similar computations. X is non-Markovian due to the part 0 f (t −
s)Xs ds in the drift which depends on the whole trajectory of X up to time t, hence the name
path-dependent.

The expression (2.4) provides essential insights to the dynamic of the process X: for small t, one
Rt
expects the non-Markovian term 0 f (t − s)Xs ds to be negligible, so that the process X behaves
locally like an Ornstein-Uhlenbeck process with large mean-reversion speed (1/2 − H)ε−1 and vol
Rt
of vol εH−1/2 for small ε. For large t, the non-Markovian term 0 f (t − s)Xs ds becomes more
prominent and introduces path-dependency.

The path-dependent Bergomi model is very different to the rough Bergomi model. It is true that
for H ∈ (0, 1/2), by setting ε = 0 in (2.3), one recovers the rough Bergomi model. However for
any ε > 0, the path-dependent Bergomi model produces a distinct model dynamic. This can be
seen from its ATM skew formula [35] below in the first order of vol of vol, in contrast to that of
the rough Bergomi model in (2.2):

(T + ε)H+3/2 εH+3/2
 
ηρ H+1/2
ST ≈ − −ε T .
2(H + 1/2)T 2 H + 3/2 H + 3/2

This formula shows that the global shape of the ATM skew of the path-dependent Bergomi model
is more flexible than the power-law shape of the rough Bergomi model. For very short maturities,
H−1/2
the ATM skew of the path-dependent Bergomi model approaches a finite limit ρηε 4 which
can be made as arbitrarily large as necessary via different values of ε, in contrast to the blow up
to ∞ in the rough Bergomi model. For longer maturities, the ATM skew of the path-dependent
Bergomi model decays at a rate ∼ ρηT H−1/2 , with the crucial difference that H can be negative,
thus allowing the ATM skew to decay faster than that of rough Bergomi model.

7
For this paper, we fix the value ε = 1/52 without calibration. This choice not only ensures the
same number of calibratable parameters as the rough Bergomi model, but also underscores that
the flexibility of the path-dependent model comes from eliminating the rough, non-semimartingale
behavior (i.e. removing the singularity of the kernel at K(0)) of rough volatility models and not
from the calibration of an additional parameter, as we will see in Section 4.

2.3 The one-factor Bergomi

The one-factor Bergomi model introduced in [12] and [24] models X using a standard Ornstein
Uhlenbeck process: Z t
−1
Xt = ηεH−1/2 e−(1/2−H)ε (t−s) dWs ,
0
where H ∈ (−∞, 1/2]. For small ε and H, X has a large mean reversion speed of order (1/2−H)ε−1
and a large vol of vol of order εH−1/2 . This way of parameterizing X is reminiscent of models
of fast regimes in [2, 4, 7, 26, 46] and can be seen as a Markovian proxy of the path-dependent
Rt
Bergomi model from the previous section by dropping the non-Markovian term 0 f (t − s)Xs ds in
(2.4).

This way of parametrization allows H to take on a similar interpretation to that in the path-
dependent Bergomi model: the more negative the H, the statistically rougher the sample path of
X driven by larger mean reversion and vol-of-vol, see Section 5.2 below. It also allows us to easily
compare the models since they have the same calibratable parameters.

The ATM skew produced by the one-factor Bergomi model at first order of vol of vol, assuming a
flat ξ0 is of the from [14]:
H−1/2
ρ εH+1/2 η  (1 − e ε T )ε 
ST ≈ 1− ,
2 (1/2 − H)T (1/2 − H)T
H−1/2
which shares the same finite limit ρηε 4 to that of the path-dependent Bergomi model. For
small T , the ATM skew decay for both models are similar: a simple computation reveals that both
models have the same limit in their first order derivative in T with respect to the skew:
dST ηρ(H − 1/2)εH−3/2
lim =
T−→0 dT 12

Therefore we can expect similar model behavior between the path-dependent and one-factor
Bergomi models for very short maturities, see Section 4.1. On the other hand, the ATM skew
decay of the one-factor Bergomi model is ∼ 1/T for large T .

2.4 The (under-parameterized) two-factor Bergomi

The two factor Bergomi model introduced in [12] contains two Ornstein Uhlenbeck factors X 1 and
X 2 in the instantaneous variance V :

1 2
Xt = Xt + Xt R,

t −1
Xt1 = ηεH−1/2 0 e−(1/2−H)ε (t−s) dWs ,
t −1
Xt = ηℓ εHℓ −1/2 0 e−(1/2−Hℓ )ε (t−s) dWs ,

 2 R

where the parametrization of both factors are derived in the same way as the one-factor Bergomi
model above with (H, Hℓ ) ∈ (−∞, 1/2]2 .

In the literature, the two factors are usually driven by two correlated Brownian motions. This
would require two extra parameters to model the correlation between the factors themselves and

8
with the Brownian motion B in the spot process S in (2.1). In practice, the calibrated correlation
between X 1 and X 2 are often very high, and for the sake of comparability and fairness among
the models, we will use the same Brownian motion W for both factors, i.e. V is Markovian in
two-dimension (X 1 , X 2 ) driven by a single Brownian motion W .

By fixing ε = 1/52, Hℓ = 0.45, we induce a fast factor X 1 (with large mean-reversion and large vol
of vol for small H) and a slow factor X 2 to mimic different scaling of volatility similar to the path-
dependent Bergomi model without sacrificing the Markovian property. On the ATM skew, the two
factors can decouple the short and long end of the term structure, with the fast factor exerting
more influence on the short end and the slow factor becoming more important as T increases.
Indeed, the ATM skew assuming a flat ξ0 at first order of vol of vol of the two-factor (or any N
factor) Bergomi model is linear in the contribution of each individual factor to the ATM skew [14]:
Hℓ −1/2
H−1/2
!
ρ εH+1/2 η  (1 − e ε T )ε  εHℓ +1/2 ηℓ  (1 − e ε T )ε 
ST ≈ 1− + 1− ,
2 (1/2 − H)T (1/2 − H)T (1/2 − Hℓ )T (1/2 − Hℓ )T

H−1/2 Hℓ −1/2
where ST has a finite limit ρ(ηε +ηℓ ε
4
)
when sending T − → 0. For maturities up to three
H−1/2
years, the two factor Bergomi models can mimic ∼ T power law decay [13, 35], despite of
having the same asymptotic for very large T as the one-factor Bergomi model.

3 Model Assessment

Our empirical study involves calibrating each model described in Section 2 to the daily SPX
volatility surfaces between August 2011 to September 2022 with market data purchased from the
CBOE website, https://datashop.cboe.com/. In total, there are 2,807 days of SPX implied
volatility surfaces.

To date, there are no closed form formulae for pricing vanilla options for the family of Volterra
Bergomi models. To speed up the tedious numerical optimization procedure and to ensure a fair
comparison, we rely on the generic-unified method ‘deep pricing with quantization hints’ developed
in [7]. This method allows us to price vanilla derivatives efficiently and accurately by combining
functional quantization and Neural Networks. In a nutshell, functional quantization is first used to
obtain a decent approximation of the vanilla option price, with Neural Network trained on Monte
Carlo prices added on later as a corrector. This approach is versatile enough to be applied to all
models where volatility is a function of Gaussian process. The main advantages of this method are:
1) it is free from butterfly arbitrage by construction and mesh-free, allowing one to price any strike
and maturity combination without interpolation or extrapolation, and 2) lower Neural Network
input dimension allowing generalization over a wide range of forward variance curve ξ0 (·). Please
refer to [7, Section 5] for detailed implementation.

3.1 Treatment of the forward variance curve

We use the same forward variance curve ξ0 (·) across all four models inferred directly from CBOE
option prices via the well-known log-contract replication formula [16]. We further assume that
ξ0 (·) is a piece-wise constant càdlàg function as suggested by Lorenzo Bergomi himself in [13]:
N
X
ξ0 (t) = 1t∈[Ti ,Ti+1 ) ξi ,
i=1

9
where Ti are available SPX option maturities, T0 := 0 and ξi > 0. We can extract ξi via:
Z S0 Z +∞ !
P (K, Ti+1 ) C(K, Ti+1 )
(Ti+1 − Ti )ξi = 2 dK + dK (3.1)
0 K2 S0 K2
Z S0 Z +∞ !
P (K, Ti ) C(K, Ti )
−2 dK + dK ,
0 K2 S0 K2

where C(K, Ti ) and P (K, Ti ) are the market prices of vanilla call/put options with strike K and
maturity Ti . Due to the scarcity of market price for deep out of money options especially for neg-
ative log moneyness, we interpolated each slice of SPX smile using a pre-determined methodology
(e.g. SSVI) then proceed with the computation in (3.1). Note we are only using the interpolated
surface to estimate ξ0 , the actual calibration is performed using the CBOE data.

3.2 Calibration performance metrics

For model evaluation, we investigate both the static performance of the model measured by how
accurate it fits the global SPX smiles and the ATM skew, and the dynamical performance of the
model, measured by how well it predicts future SPX smiles.

3.2.1 Global fit of the implied volatility surface

To test the global fit of the SPX smiles, each model is calibrated by minimizing the error function
between model implied volatility surface and that of SPX over the set of calibratable model param-
eters, denoted collectively as Θ. For the rough, path-dependent and one-factor Bergomi models,
Θ = {η, ρ, H}. For the two-factor Bergomi model, Θ = {η, ρ, H, ηℓ }. We chose the Root Mean
Square Error (RMSE) as the error function J (Θ):
v
u  2
u 1 X mid
J (Θ) := t bi,j − σ
σ bi,j (Θ) , (3.2)
|I|
(i,j)∈I

mid
with σbi,j the SPX mid implied volatility with maturity Ti and strike Kj , and σbi,j (Θ) the model
implied volatility. The set of available SPX implied volatility data for different maturities and
strikes is captured by the index set I, with |I| denoting the total number of available data points.

Due to the availability of market data as well as the stability of the implied volatility estimator,
we use the following log-moneyness (k = log(K/S0 )) range shown in Table 2 for model calibration:

Maturities T log moneyness range

< 2 weeks [−0.15, 0.03]
< 1 month [−0.25, 0.03]
< 2 months [−0.3, 0.04]
< 3 months [−0.4, 0.15]
< 6 months [−0.6, 0.15]
< 1 year [−0.8, 0.2]
≥ 1 year [−1.5, 0.3]

Table 2: Log moneyness range for different maturities considered for model calibration to the
global SPX volatility surface.

10
3.2.2 Fit of the implied volatility ATM skew

To evaluate model fit of the ATM skew, each model is calibrated to the SPX ATM skew by
minimizing the error between model implied volatility and the mid SPX implied volatility over a
narrow range of log moneyness k ∈ [−0.05, 0.03] across all maturities, as well as the error between
log of model ATM skew and log of SPX ATM skew:
(v )
u 1 X  mid
u 2 1 X  2
min t bi,j − σ
σ bi,j (Θ) + mkt
log SbTm − log SbTm (Θ) ,
Θ |I| |M|
(i,j)∈I (m)∈M

where SbTmkt
m
is the SPX ATM skew at maturity Tm , with SbTm (Θ) is the model ATM skew. The set
of available market ATM skew for different maturities is captured by the index set M.

Thanks to the mesh-free nature of our deep pricing method that combines Functional Quantization
and NN, we are able to calibrate each model directly to the SPX ATM skew without resorting
to the approximation formula of the Bergomi-Guyon expansion as is the case in [22, 38]. While
the Bergomi-Guyon expansion provides a fast approximation of model ATM skew, it can be error-
prone in the case involving rough volatility [10] and situations where the vol of vol is large in
Markovian-factor models.

To compute the ATM skew of each model, we first compute the model implied volatility near the
money and then use the central finite difference at k = 0. The SPX ATM skew is computed by
fitting a Lagrange polynomial of order 3 locally near the money and then taking the first order
derivative at k = 0. We checked our results to ensure no over/under-fitting.

3.2.3 Prediction quality (parameter stability)

This performance metric tests the dynamical performance of each model by looking at their pre-
diction quality. The main idea is that a robust parametric model should not require frequent
re-calibration, since the change in the state variables (i.e. ξ0 (·)) alone should be sufficient in jus-
tifying the change in the output (i.e. deformation in the implied volatility surface). Thus this
performance metric can also be seen as a test on the stability of calibrated parameters.

To evaluate the prediction quality of a model, we perform the following experiment: for each
trading day, we take the calibrated parameters Θ∗ obtained by minimizing the error function (3.2)
and keep it fixed for the next 20 working days. Next, we take the daily market forward variance
curve ξ0 (·) extracted as per (3.1) and compute the error function (3.2) for each of the next 20
working days using Θ∗ .

This is similar to the test described in Section 4.2 from [50]. However, the treatment of ξ0 (·) in
[50] is not consistent for every model: the ξ0 (·) for the Markovian Heston model is modelled by
only three parameters (V0 , κ, V∞ ), where the rough Bergomi and rough Heston models received
preferential treatment by employing piece-wise constant function between maturities that offers
greater flexibility in controlling the overall level of the model implied volatility. In addition, ξ0 (·)
in [50] is re-calibrated daily for each model. In our experiment, we do not modify the market ξ0 (·)
to ensure the state variable is being respected as much as possible.

4 Empirical results: the static comparison

This section on empirical performance is concentrated on the static performance of each model and
is divided into two parts. The first part concentrate on model performance over the short term
with maturities between one week to three months. The second part is dedicated to a larger range

11
of maturities spanning between one week to three years. We will defer the discussion on dynamical
performance of each model to Section 5.

4.1 Short maturities (one week to three months)

We compare the calibration performance between rough, path-dependent and one-factor Bergomi
models that share the same number of calibratable parameters. Figure 1 shows the evolution of
daily calibration RMSE (3.2) of global fits of the implied volatility surface by each model. For
ease of comparison, we show the monthly rolling average RMSE. The calibrated error for the
path-dependent Bergomi model is almost always below that of the rough Bergomi model, while
the one-factor Bergomi model also outperforms the rough Bergomi model about 63 percent of the
time. The summary of statistics in Table 3 shows that the rough Bergomi model also scores the
highest variance and the worst performance across the board. Some example fits are provided in
Appendix A.

Figure 1: Evolution of monthly rolling average of calibration RMSE between different Bergomi
models for maturities between one week to three months. All models share the same number of
calibratable parameters, namely (η, ρ, H).

Model mean std min 5% 50% 95% max

rough 0.005483 0.001786 0.001059 0.002996 0.005277 0.008661 0.01984
path-dependent 0.004859 0.001550 0.000937 0.002739 0.004697 0.007523 0.018898
one-factor 0.004987 0.001518 0.001070 0.002844 0.004811 0.007615 0.016218

Table 3: Statistics on the calibration RMSE for maturities between one week to three months.
The lowest error for each statistical measure is in bold.

Figure 2 shows different model fits to the SPX ATM skew by plotting the average ATM skew
and log of average ATM skew. The calibrated ATM skew from the one-factor Bergomi model is
the closest to the market data for maturities up to one and half months, followed by the path-
dependent Bergomi model. Even though no model seems to be able to capture perfectly the entire
term structure of the ATM skew under three months, the rough Bergomi model failed to capture
the market ATM skew at any point, characterized by an overly-explosive ATM skew near one week
and a much too rapid decay straight after. We thus have evidence that rough volatility models are

12
inconsistent with market data at short timescales between one week and three months, and that
ATM skew cannot be adequately explained by the power law generated by rough volatility models.

Figure 2: Average SPX ATM skew term structure (left) and log average SPX ATM skew term
structure in log-scale (right) produced by different Bergomi models with maturities calibrated
between one week to three months. We zoomed in for maturities between one to two weeks.

Contrary to what rough volatility literature claims, we found no evidence supporting that
rough volatility models fit better than their Markovian counterparts for short ma-
turities of the SPX smile. Instead, our study shows that the one-factor Bergomi actually
outperforms the rough Bergomi model, which is in line with our earlier paper on joint SPX-VIX
calibration [7, 8] where the Quintic one-factor Markovian model is shown to consistently outperform
its rough counterpart for maturities up to 3 months.

Since all these models use the same calibratable parameters (η, ρ, H), we can compare their evo-
lution in time. Figure 3 shows the calibrated ρ for the rough Bergomi model tends to saturate
near −1 (and even more so in Section 4.2 for longer maturities): this is a known structural issue
of the rough Bergomi model, see [7, 25, 42, 50]. This saturation cannot be explained by the under-
parametrization of the rough Bergomi model (since all the models here have the same number of
parameters), but rather by its inconsistency with the market observation. Indeed, the other two
models have very similar behavior in their calibrated parameters, which highlights again that the
path-dependent model is very different from the rough model. Figure 4 shows that the calibrated
H for the path-dependent and one-factor Bergomi models are almost always negative, versus that
of the rough Bergomi model. The negative H ≤ −1/2 can be related to jump regimes at the limit
when ε −→ 0, under Heston’s dynamics [2].

13
Figure 3: Evolution of monthly average of calibrated parameter ρ for different models calibrated
to maturities between one week to three months.

Figure 4: Evolution of monthly average of calibrated parameter H for different models calibrated
to maturities between one week to three months.

4.2 Long maturities (one week to three years)

We now compare the empirical results for maturities running all the way to three years for the
rough, path-dependent, one-factor and under-parameterized two-factor Bergomi models. We recall
the two-factor Bergomi model contains one-extra calibratable parameter ηℓ compared to the other
models but with both factors driven by the same Brownian motion.

Figure 5 and Table 4 show the two-factor Bergomi model outperforms all other models for the
global fit of the implied volatility surface, followed by the path-dependent Bergomi model. The
performance of the rough Bergomi model is noticeably unsatisfactory. It scores again the highest
variance. Even though its performance is slightly better than that of the one-factor Bergomi
model (around 57.4 percent of the time), for the period between 2017 to 2019, it consistently
underperformed the one-factor Bergomi model. Some example fits are provided in Appendix B.

14
Figure 5: Evolution of monthly rolling average of calibration RMSE between different Bergomi
models for maturities between one week to three years. Recall rough, path-dependent and one-
factor Bergomi models share the same number of calibratable parameters (η, ρ, H) while the two-
factor has an extra calibratable parameter ηℓ .

Model mean std min 5% 50% 95% max

rough 0.007725 0.002113 0.002904 0.004755 0.007394 0.011517 0.020934
path-dependent 0.005935 0.001613 0.002271 0.003695 0.005755 0.008729 0.021469
one-factor 0.007943 0.001905 0.003287 0.005133 0.00781 0.011234 0.022543
two-factor 0.005431 0.001429 0.002403 0.003554 0.005272 0.007766 0.021952

Table 4: Statistics on the calibration error for the maturities between one week to three years.
The lowest error for each statistical measure is in bold.

From Figure 6, the two-factor and the path-dependent Bergomi models produce on average a
decent fit, although not perfect to the full term structure of the SPX ATM skew. Both the rough
Bergomi and the one-factor Bergomi models produce equally bad fit and are inconsistent with the
general shape of SPX ATM skew, although the one-factor Bergomi model seems to at least be able
to achieve a curved shape in log-log plot more representative of the market vs. the straight line of
rough volatility models. Like [22, 38], we now have further evidence challenging the assumption
used by rough volatility advocates that the SPX ATM skew follows a power law. The straight
line fit of the rough Bergomi model in the log-log plot on the right hand side is in stark contrast
with the curved shape of market average log ATM skew term structure. In addition, its explosive
behavior for small maturities results in overestimation for the short end; and for the long end the
rough Bergomi model is unable to match the speed of decay of the market.

15
Figure 6: Average SPX ATM skew term structure (left) and log average SPX ATM skew term
structure in log-scale (right) produced by different Bergomi models for maturities between one
week to three years. Recall rough, path-dependent and one-factor Bergomi models share the
same number of calibratable parameters (η, ρ, H) while the two-factor has an extra calibratable
parameter ηℓ .

Our results provides strong evidence against the claim that rough volatility models con-
sistently outperform their classical Markovian counterparts. First, it is not clear that
the rough Bergomi model actually performs better than the one factor Bergomi model. Second,
the under-parameterized two-factor Bergomi model consistently outperforms the rough Bergomi
model only with one additional parameter ηℓ . As observed in [22, 50, 36], having two rough factors
driving the volatility process in the Bergomi model does not significantly improve the model fit
compared to the standard two factor (Markovian) Bergomi model.

Figure 7 shows that the calibrated ρ for the rough Bergomi model tends to saturate at −1, making
it somewhat a redundant parameter. The calibrated ρ for all other models are not saturated and
move similar to one another:

Figure 7: Evolution of monthly average of calibrated parameter ρ of different Bergomi models

calibrated to maturities between one week to three years.

The calibrated H shown in Figure 8 for the path-dependent and two-factor Bergomi models are
negative, in contrast with that of the rough Bergomi model. The H from the one-factor Bergomi
model is usually the highest, but dipped below zero regularly for the period between 2017 and

16
2019. This coincides with the same time-frame where the one-factor Bergomi model outperformed
the rough Bergomi model shown in Figure 5. Compared to Figure 4, the calibrated H here is
about 0.4 higher for the path-dependent Bergomi model: the increase in H is necessary to help
the model taking care of the longer term maturities, since the more negative the H, the faster the
ATM skew decay for larger T .

Figure 8: Evolution of monthly average of calibrated parameter H of different Bergomi models

calibrated to maturities between one week to three years.

4.3 What does negative H mean?

On the short end, the large negative values of H can be difficult to interpret. For small maturities,
the large negative calibrated values of H observed in the path-dependent and one-factor model
on Figure 4 can be attributed to a highly erratic/spiky behavior of the volatility process, see
Appendix C, which might not be well explained by a rough process, but may indicate the presence
of jumps. Indeed, negative H ≤ −1/2 can be related to jump regimes at the limit when ε − →0
under Heston’s dynamics [2]. These negative values for H also appeared in Figure 3 and Appendix
B of [7] for the problem of joint calibration of SPX and VIX. When calibrated on longer maturities,
the path-dependent model still displays a negative calibrated H as shown in Figure 8, although
less extreme. This can be more easily explained by looking at the ATM skew as follows.

In two recent independent studies [22, 38], two different power laws have been used to fit the short
and long end of the SPX ATM skew. We perform the same experiment by fitting two power laws
both in the form of cT H−1/2 on the average SPX ATM skew using linear regression. Specifically,
we fit one power law on (log T, log S̄Tspx ), with S̄Tspx the average SPX ATM skew over the period
for T < τ and another power law for T ≥ τ and infer the two different value of H at different
timescales. In Figure 9, τ is chosen to be 4 months based on the highest average R2 value of the
two linear regressions among all possible values of τ between one week to three years:

17
Figure 9: Fitting the log S̄Tspx with two linear regressions on log T , with the first linear regression
performed on T ∈ [1W, 4M) and the second linear regression performed on T ∈ [4M, 3Y]. The
slope of each linear regression represents estimated H at different timescales.

The decent fit of the blue dash line in Figure 9 suggests that long term ATM skew can be well-
captured by a power law but with a negative H2 = −0.095. We now show what happens to the
estimation of H2 by moving the cut-off time τ between one month to one year:

Figure 10: Fitting the log S̄Tspx with linear regression on log T for T ∈ [τ, 3Y] with τ ∈
{1M,2M,4M,1Y}. The slope of each linear regression represents estimated H2 at different
timescales.

The estimated H2 becomes more negative as τ increases. This suggests that on average, the SPX
ATM skew decays even faster than previously reported in [13, 28, 31, 32], where it is believed the
decay coefficient is greater than −1/2, i.e. H ≥ 0. The restriction H > 0 in the rough Bergomi
model makes it impossible to match the average speed of SPX ATM skew decay for large T and
partially explains its poor fit to the long end of the term structure of the SPX ATM skew in Figure
6. Similar negative values of H2 have also appeared in Figure 4.8 of [37] when fitting the ATM
skew on a specific day; and in Appendix C of [22] for fitting the average log ATM skew of a different
and narrower time window.

18
To further validate the negative H2 observed in Figures 9 and 10, we fit the daily SPX log ATM
skew term structure against log T for T ∈ [1Y, 3Y] via linear regression and plot the evolution of
H2 in Figure 11 with calibrated H from the rough and path-dependent Bergomi models (calibrated
to the ATM skew for maturities between one week to three years) added in for comparison:

Figure 11: Evolution of market H2 estimated by fitting the daily log SPX ATM skew against log T
for T ∈ [1Y, 3Y] compared to calibrated value of H of the rough and path-dependent Bergomi
models (calibrated to the ATM skew for maturities between one week to three months). The dotted
line represents the value zero.

Figure 11 shows that the estimated market H2 is almost always negative apart from 2012 to mid
2013, thus providing further support to the case of negative H for SPX long term ATM skew decay.

Why should one care about the precise long term decay of the ATM skew? The long term ATM
skew decay is linked to dynamical properties of the model, for instance via the skew stickiness ratio
(SSR), where ST ∝ T SSR−2 for large T [13]. Estimating the SSR could then help in deducing
H2 through the relation H2 ≈ SSR − 1.5. In particular, the restriction of H ∈ (0, 1/2) in rough
volatility models leads to an SSR ∈ (1.5, 2). We performed a first estimation of SSR using options
with one year maturity in our data. The resulting SSR estimation fluctuates too much between
0.6 and 1.6 to obtain a reliable estimate of H2 . However, the range of estimated SSR does suggest
that H2 can be negative which is inconsistent with rough volatility models, but can be better
captured by the path-dependent and two factor models. In the next section, we will evaluate the
dynamical properties of the models using a different experiment.

5 Empirical results: the dynamical comparison

5.1 Predicting Future SPX Implied Volatility Surface

We now compare the dynamical performance of each model by looking at how well they predict
the future SPX volatility surface as described in Section 3.2.3. That is, for a given model, we fix
the daily calibrated parameters obtained while fitting the global SPX smile in Section 3.2.1. We
keep the calibrated parameters fixed for the next 20 working days, while taking in the daily market
forward variance curve ξ0 (·) and see how well the model predicts SPX volatility surface in the next
20 working days, measured by the RMSE error between model vs. market mid data.

The box plot in Figure 12 shows the distribution of RMSE of prediction quality as across the whole
period when the parameters are calibrated only to short maturities (one week to three months).
The path-dependent Bergomi model scores the best performance. The one-factor Bergomi model

19
performs better compared to the rough Bergomi model for the first six forward business days and
shares similar performance afterwards.

Figure 12: Box plot distribution of prediction RMSE between different Bergomi models across
the entire period for maturities between one week to three months. Different values on the box
plot represent 25, 50 and 75 quantiles, while the whiskers are calculated as 1.5 multiplied by the
inter-quartile range away from 25 and 75 quantiles.

The box plot in Figure 13 shows the distribution of RMSE of prediction quality as across the whole
period when the parameters are calibrated to longer maturities (one week to three years). The
two-factor Bergomi model is the best among all the models in predicting future volatility surface,
followed by the path-dependent Bergomi model. The rough Bergomi model slightly outperforms
the one-factor model, however is considerably inadequate when compared to the other models.

Figure 13: Box plot distribution of prediction RMSE between different Bergomi models across
the entire period for maturities between one week to three years. Different values on the box
plot represent 25, 50 and 75 quantiles, while the whiskers are calculated as 1.5 multiplied by the
inter-quartile range away from 25 and 75 quantiles.

Both graphs show that the dynamical evolution of rough volatility model is far less consistent

20
with market data than its path-dependent and Markovian-counterparts. This further highlights
the rigid structural issues associated with rough volatility models.

5.2 Spurious Roughness

We now use the same methodology in [33] to estimate the statistical roughness of the instanta-
neous volatility process via its proxy, the realized volatility process RV of all Bergomi models, as
instantaneous volatility
√ is unobservable in practice. First, we simulate the trajectories of the in-
stantaneous volatility V and log S under each model, with time step size of 5 minutes for T = 10
years. Next, we compute the estimated daily RV via the following:
v
u n  2
uX
RV = t log(Sti /Sti−1 ) ,
i=1

with tn − t0 = 1 day and then calculate the empirical q-variation of the form:
N q
1 X
m(q, ∆) = RVk∆ − RV(k−1)∆
N
k=1

for different values of q and timescale ∆. To remain as close as possible to the experiments
performed in [33], we choose q ∈ {0.5, 1, 1.5, 2, 3} and timescale ∆ = 1, 2, . . . , 50 Days. By fixing
q, we compute the empirical q-variation and then use linear regression on (log ∆, log m(q, ∆)) to
estimate its slope ζq . Next, we plot (q, ζq ) and the slope of the graph (q, ζq ) is the estimated Hurst
index of the simulated process RV.
√
To simulate instantaneous volatility V , we use the calibrated parameters of each model for the
implied volatility surface on the day 23 October 2017, calibrated for maturities up to three months
with the same seed. For this date, the calibrated parameters for the models are:

Model Calibrated parameters

rough η = 1.28, ρ = −0.9397, H = 0.0787
path-dependent η = 0.02551, ρ = −0.6884, H = −1.2761
one-factor η = 0.7555, ρ = −0.6842, H = −0.3639
two-factor η = 0.4297, ηℓ = 0.9843, ρ = −0.6849, H = −0.4973

Table 5: Calibrated model parameters of each model for the implied volatility surface on 23 October
2017, with calibrated maturities between one week and three years.

21
Figure 14: The log-log plot of different q-variations of a sample path of the realized volatility
process of different models.

Figure 15: Plot of ζq against q, with the slope being the estimated Hurst index of different models.

22
Figure 16: LHS: The log-log plot of different q-variations of the realized volatility time series of
S&P500 between 2007 to 2017. RHS: Plot of ζq against q, with the slope being the estimated Hurst
index of the S&P500 realized volatility time series.

From Figure 14 and 15, we see that the estimate for the Hurst index of the rough Bergomi model
b ≈ 0.128 comparing to the calibrated value of H = 0.0787 for the instantaneous volatility.
is H
For the other models, the estimated Hurst index Hb of RV returns a value between 0.097 to 0.159,
comparing to the theoretical Hurst index of 0.5 of their instantaneous volatility. Figure 16 shows
the estimated Hb ≈ 0.137 of the actual S&P500 realized volatility over a ten year period between
2007 to 2017, with data coming from the Oxford-Man Institute of Quantitative Finance Realized
Library for comparison.

The sample paths of realized volatility used for the estimation of H in Figure 14 and 15 of each
Bergomi model are included in Appendix C alongside with the annualized realized volatility time
series of the S&P500. These sample paths all seem to have an erratic and spiky behavior. Combined
with the small estimated statistical H,
b this suggests that the rougher appearance of the realized
volatility process does not mean that the underlying process has to be rough, and the estimation
of H is not enough to discriminate between models, since all of our non-rough models produce
similar spurious roughness on any realistic timescales, in line with [1, 19, 39, 49].

23
A Sample fits of SPX smile: one week to three months

July 03, 2013

Figure 17: SPX smiles (bid/ask in blue/red) on 3 July 2013 calibrated by different Bergomi models
(green lines).

October 23, 2017

24
Figure 18: SPX smiles (bid/ask in blue/red) on 23 October 2017 calibrated by different Bergomi
models (green lines).

B Sample fits of SPX smile: one week to three years

July 03, 2013

Figure 19: SPX smiles (bid/ask in blue/red) on 3 July 2013 calibrated by rough Bergomi models
(green lines).

25
Figure 20: SPX smiles (bid/ask in blue/red) on 3 July 2013 calibrated by path-dependent
Bergomi models (green lines).

Figure 21: SPX smiles (bid/ask in blue/red) on 3 July 2013 calibrated by one-factor Bergomi
models (green lines).

Figure 22: SPX smiles (bid/ask in blue/red) on 3 July 2013 calibrated by two-factor Bergomi
models (green lines).

26
October 23, 2017

Figure 23: SPX smiles (bid/ask in blue/red) on 23 October 2017 calibrated by rough Bergomi
models (green lines).

Figure 24: SPX smiles (bid/ask in blue/red) on 23 October 2017 calibrated by path-dependent
Bergomi models (green lines).

27
Figure 25: SPX smiles (bid/ask in blue/red) on 23 October 2017 calibrated by one-factor Bergomi
models (green lines).

Figure 26: SPX smiles (bid/ask in blue/red) on 23 October 2017 calibrated by two-factor Bergomi
models (green lines).

C Simulated realized volatility

Figure 27 shows the simulated trajectories of the realized volatility process under each Bergomi
model with the same seed, using the calibrated parameters from Table 5 based on the implied
volatility surface on 27 October 2017, and the annualized realized volatility time series of S&P 500
between 2007 to 2017 with data coming from the Oxford-Man Institute’s Realized Library.

28
Figure 27: Simulated trajectory of the realized volatility process RV of different models with the
same seed using the calibrated parameters of the implied volatility surface dated 27 October 2017
and the annualized realized volatility time series of S&P 500 between 2007 to 2017.

References
[1] Eduardo Abi Jaber. Lifting the Heston model. Quantitative Finance, 19(12):1995–2013, 2019.
[2] Eduardo Abi Jaber and Nathan De Carvalho. Reconciling rough volatility with jumps. Avail-
able at SSRN 4387574, 2023.
[3] Eduardo Abi Jaber and Omar El Euch. Markovian structure of the Volterra Heston model.
Statistics & Probability Letters, 149:63–72, 2019.
[4] Eduardo Abi Jaber, Martin Larsson, and Sergio Pulido. Affine Volterra processes. The Annals
of Applied Probability, 29(5):3155–3200, 2019.
[5] Eduardo Abi Jaber, Christa Cuchiero, Martin Larsson, and Sergio Pulido. A weak solution

29
 theory for stochastic volterra equations of convolution type. The Annals of Applied Probability,
31(6):2924–2952, 2021.
[6] Eduardo Abi Jaber, Enzo Miller, and Huyên Pham. Linear-quadratic control for a class of
stochastic volterra equations: solvability and approximation. The Annals of Applied Proba-
bility, 31(5):2244–2274, 2021.
[7] Eduardo Abi Jaber, Camille Illand, and Shaun Xiaoyuan Li. Joint SPX–VIX calibration
with gaussian polynomial volatility models: deep pricing with quantization hints. Available
at SSRN 4292544, 2022.
[8] Eduardo Abi Jaber, Camille Illand, and Shaun Li. The quintic ornstein-uhlenbeck volatility
model that jointly calibrates spx & vix smiles. Risk Magazine, Cutting Edge Section, 2023.
[9] Elisa Alos, Jorge A León, and Josep Vives. On the short-time behavior of the implied volatility
for jump-diffusion models with stochastic volatility. Finance and stochastics, 11(4):571–589,
2007.
[10] Christian Bayer, Peter Friz, and Jim Gatheral. Pricing under rough volatility. Quantitative
Finance, 16(6):887–904, 2016.
[11] Christian Bayer, Peter K Friz, Paul Gassiat, Jorg Martin, and Benjamin Stemper. A regularity
structure for rough volatility. Mathematical Finance, 30(3):782–832, 2020.
[12] L Bergomi. Smile dynamics II. Risk Magazine, 2005.
[13] Lorenzo Bergomi. Stochastic volatility modeling. CRC press, 2015.
[14] Lorenzo Bergomi and Julien Guyon. Stochastic volatility’s orderly smiles. Risk, 25(5):60,
2012.
[15] Ofelia Bonesini, Antoine Jacquier, and Alexandre Pannier. Rough volatility, path-dependent
pdes and weak rates of convergence. arXiv preprint arXiv:2304.03042, 2023.
[16] Peter Carr and Dilip Madan. Towards a theory of volatility trading. Option Pricing, Interest
Rates and Risk Management, Handbooks in Mathematical Finance, 22(7):458–476, 2001.
[17] Peter Carr and Liuren Wu. What type of process underlies options? a simple robust test.
The Journal of Finance, 58(6):2581–2610, 2003.
[18] Fabienne Comte and Eric Renault. Long memory in continuous-time stochastic volatility
models. Mathematical finance, 8(4):291–323, 1998.
[19] Rama Cont and Purba Das. Rough volatility: fact or artefact? arXiv preprint
arXiv:2203.13820, 2022.
[20] Sylvain Corlay, Joachim Lebovits, and Jacques Lévy Véhel. Multifractional stochastic volatil-
ity models. Mathematical Finance, 24(2):364–402, 2014.
[21] Christa Cuchiero and Josef Teichmann. Generalized feller processes and markovian lifts of
stochastic volterra processes: the affine case. Journal of evolution equations, 20(4):1301–1348,
2020.
[22] Jules Delemotte, Stefano De Marco, and Florent Segonne. Yet another analysis of the sp500
at-the-money skew: Crossover of different power-law behaviours. Available at SSRN 4428407,
2023.
[23] Zhuanxin Ding, Clive WJ Granger, and Robert F Engle. A long memory property of stock
market returns and a new model. Journal of empirical finance, 1(1):83–106, 1993.
[24] Bruno Dupire. Arbitrage pricing with stochastic volatility. Société Générale, 1992.
[25] Martin Forde, Masaaki Fukasawa, Stefan Gerhold, and Benjamin Smith. The rough bergomi
model as h→ 0–skew flattening/blow up and non-gaussian rough volatility. preprint, 2020.

30
[26] Jean-Pierre Fouque, George Papanicolaou, and K Ronnie Sircar. Derivatives in financial
markets with stochastic volatility. Cambridge University Press, 2000.

[27] Jean-Pierre Fouque, George Papanicolaou, Ronnie Sircar, and Knut Solna. Multiscale stochas-
tic volatility asymptotics. Multiscale Modeling & Simulation, 2(1):22–42, 2003.
[28] Jean-Pierre Fouque, George Papanicolaou, Ronnie Sircar, and Knut Solna. Maturity cycles
in implied volatility. Finance and Stochastics, 8:451–477, 2004.

[29] Peter K Friz, Jim Gatheral, and Radoš Radoičić. Forests, cumulants, martingales. The Annals
of Probability, 50(4):1418–1445, 2022.
[30] Masaaki Fukasawa. Asymptotic analysis for stochastic volatility: martingale expansion. Fi-
nance and Stochastics, 15:635–654, 2011.

[31] Masaaki Fukasawa. Volatility has to be rough. Quantitative Finance, 21(1):1–8, 2021.
[32] Jim Gatheral. Consistent modeling of spx and vix options. In Bachelier congress, volume 37,
pages 39–51, 2008.
[33] Jim Gatheral, Thibault Jaisson, and Mathieu Rosenbaum. Volatility is rough. Quantitative
finance, 18(6):933–949, 2018.

[34] Archil Gulisashvili. Large deviation principle for volterra type fractional stochastic volatility
models. SIAM Journal on Financial Mathematics, 9(3):1102–1136, 2018.
[35] Julien Guyon. The smile of stochastic volatility: Revisiting the bergomi-guyon expansion.
Available at SSRN 3956786, 2021.

[36] Julien Guyon. Dispersion-constrained martingale schrödinger bridges: Joint entropic calibra-
tion of stochastic volatility models to s&p 500 and vix smiles. Available at SSRN 4165057,
2022.
[37] Julien Guyon. The vix future in bergomi models: Fast approximation formulas and joint
calibration with s&p 500 skew. SIAM Journal on Financial Mathematics, 13(4):1418–1485,
2022.
[38] Julien Guyon and Mehdi El Amrani. Does the term-structure of equity at-the-money skew
really follow a power law? Available at SSRN 4174538, 2022.
[39] Julien Guyon and Jordan Lekeufack. Volatility is (mostly) path-dependent. Volatility Is
(Mostly) Path-Dependent (July 27, 2022), 2022.
[40] Philipp Harms and David Stefanovits. Affine representations of fractional processes with
applications in mathematical finance. Stochastic Processes and their Applications, 129(4):
1185–1228, 2019.
[41] Blanka Horvath, Antoine Jacquier, and Aitor Muguruza. Functional central limit theorems
for rough volatility. arXiv preprint arXiv:1711.03078, 2017.
[42] Blanka Horvath, Aitor Muguruza, and Mehdi Tomas. Deep learning volatility: a deep neural
network perspective on pricing and calibration in (rough) volatility models. Quantitative
Finance, 21(1):11–27, 2021.

[43] Antoine Jacquier and Alexandre Pannier. Large and moderate deviations for stochastic
volterra systems. Stochastic Processes and their Applications, 149:142–187, 2022.
[44] Thibault Jaisson and Mathieu Rosenbaum. Rough fractional diffusions as scaling limits of
nearly unstable heavy tailed hawkes processes. The Annals of Applied Probability, 26(5):
2860–2882, 2016. ISSN 10505164.

31
[45] Yanhui Liu, Pierre Cizeau, Martin Meyer, C-K Peng, and H Eugene Stanley. Correlations in
economic time series. Physica A: Statistical Mechanics and its Applications, 245(3-4):437–440,
1997.
[46] Serguei Mechkov. Fast-reversion limit of the Heston model. Available at SSRN 2418631, 2015.
[47] Jean-François Muzy, Jean Delour, and Emmanuel Bacry. Modelling fluctuations of financial
time series: from cascade process to stochastic volatility model. The European Physical Journal
B-Condensed Matter and Complex Systems, 17:537–548, 2000.

[48] Josep Perelló, Jaume Masoliver, and Jean-Philippe Bouchaud. Multiple time scales in volatility
and leverage correlations: a stochastic volatility model. Applied Mathematical Finance, 11(1):
27–50, 2004.
[49] L Rogers. Things we think we know, 2019.

[50] Sigurd Emil Rømer. Empirical analysis of rough and classical stochastic volatility models to
the spx and vix markets. Quantitative Finance, 22(10):1805–1838, 2022.
[51] Frederi Viens and Jianfeng Zhang. A martingale approach for fractional brownian motions
and related path dependent pdes. The Annals of Applied Probability, 29(6):3489–3540, 2019.

32