Paper: arXiv 2512.02352

Authors: Michał Sikorski

Abstract

Volatility clustering is one of the most robust stylized facts of financial markets, yet it is typically detected using moment-based diagnostics or parametric models such as GARCH. This paper shows that clustered volatility also leaves a clear imprint on the time-reversal symmetry of horizontal visibility graphs (HVGs) constructed on absolute returns in physical time. For each time point, we compute the maximal forward and backward visibility distances, \(L^{"+"}(t)\) and \(L^{"-"}(t)\), and use their empirical distributions to build a visibility-asymmetry fingerprint comprising the Kolmogorov–Smirnov distance, variance difference, entropy difference, and a ratio of extreme visibility spans. In a Monte Carlo study, these HVG asymmetry features sharply separate volatility-clustered GARCH(1,1) dynamics from i.i.d.\ Gaussian noise and from randomly shuffled GARCH series that preserve the marginal distribution but destroy temporal dependence; a simple linear classifier based on the fingerprint achieves about 90% in-sample accuracy. Applying the method to daily S&P500 data reveals a pronounced forward–backward imbalance, including a variance difference \(Δ\mathrm{“Var”}\) that exceeds the simulated GARCH values by two orders of magnitude and vanishes after shuffling. Overall, the visibility-graph asymmetry fingerprint emerges as a simple, model-free, and geometrically interpretable indicator of volatility clustering and time irreversibility in financial time series.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper introduces moderately advanced graph theory and statistical measures (e.g., Kolmogorov-Smirnov, entropy) but lacks implementation details or robust backtesting beyond a single empirical application, placing it between theoretical modeling and limited empirical validation.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Structural Indicator of<br>Volatility Clustering"] --> B["Methodology:<br>Horizontal Visibility Graphs on<br>Absolute Returns"]
  B --> C["Compute Asymmetry Metrics<br>KS Distance / Var Diff / Entropy Diff"]
  C --> D["Simulation:<br>GARCH(1,1) vs. i.i.d. vs. Shuffled Data"]
  D --> E["Application:<br>Real S&P 500 Daily Data"]
  E --> F["Key Outcome:<br>Model-Free Geometric Indicator<br>Identifies Volatility Clustering"]
  F --> G["Key Finding:<br>Imbalance vanishes after shuffling<br>Proves Temporal Dependence"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper introduces the forward visibility horizon L+(t) as a rank‑invariant, parameter‑free estimator of path roughness and uses it to distinguish rough volatility models.

Source sentence
This article introduces L+(t), the forward visibility horizon at node t, with finite-sample terminal non-crossings treated as right-censored observations.

Assumes 3

  • Persistence probability of the RL‑fBm increment process below zero scales with the same exponent as fBm of the same Hurst index. model
    Source sentence
    Suppose, in addition, that the persistence probability of the RL-fBm increment process below zero has the same scaling exponent as that of fBm of the same Hurst index, that is, …
  • The underlying time series has no ties, so uncensored L+(t) equals the first‑passage time τ+(t). model
    Source sentence
    For paths without ties, each uncensored L+(t) is identical to the first-passage time τ+(t) = inf{k ≥ 1 : x_{t+k} ≥ x_t}.
  • The data are an i.i.d. sequence with a continuous distribution for the survival law derivation. statistical
    Source sentence
    For an i.i.d. sequence with a continuous distribution, the survival law is exactly Pr[L+ ≥ k] = 1/k.

Data

  • Daily FRED VIX · FRED macro public
    2000–2026
    Source sentence
    Applied to daily FRED VIX data from 2000–2026, the rolling estimate is ˆθ = 0.91 ± 0.19 across 45 four-year windows
  • Monte‑Carlo synthetic paths synthetic public
    Source sentence
    In Monte‑Carlo experiments (N = 10,000, T = 2^16), a Hill‑MLE with Clauset–Shalizi–Newman threshold selection recovers θ(H) within one cross‑replicate standard deviation for H ≤ 0.2
  • Synthetic i.i.d. paths for null distribution synthetic public
    Source sentence
    1000 i.i.d. paths of length 6651 were also simulated and evaluated with the same rolling scheme

Recognised sources: FRED Cboe

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • The theoretical prediction θ(H)=1−H holds for rough Bergomi volatility under the persistence hypothesis. theoretical
    Source sentence
    Combining the identity L+ = τ+ with discrete‑grid persistence theory for fractional Brownian motion gives the prediction θ(H) = 1 − H.
  • Monte‑Carlo simulations recover the predicted θ(H) within one standard deviation for H ≤ 0.2, with a positive finite‑size bias for smoother paths. statistical
    Source sentence
    a Hill‑MLE with Clauset–Shalizi–Newman threshold selection recovers θ(H) within one cross‑replicate standard deviation for H ≤0.2 and reveals a positive finite‑size bias for smoother paths.
  • The rank‑invariant estimator cleanly separates rough Bergomi volatility from classical Heston, GARCH, and FIGARCH benchmarks. performance
    Source sentence
    The rank‑invariant, parameter‑free estimator separates rough Bergomi volatility from classical Heston, GARCH, and FIGARCH benchmarks.
  • Rolling VIX estimate θ̂ = 0.91 ± 0.19 is far below the i.i.d. Monte‑Carlo null (p < 0.001), indicating rough volatility in market data. statistical
    Source sentence
    the rolling estimate is ˆθ = 0.91 ± 0.19 across 45 four-year windows and lies far below an overlapping-window i.i.d. Monte‑Carlo null (p < 0.001).
  • In the VIX rolling windows, 35 out of 45 fall below the single‑window 2.5% null quantile, a result highly unlikely under the iid null (p ≃ 0.001). statistical
    Source sentence
    In the VIX data, 35/45 windows fall below the single‑window 2.5% null quantile 1.017.

Checked by others

Supported by

  • Rough volatility: evidence from range volatility estimators (2023)
    On the question: Do market volatility measures exhibit rough volatility behavior?
    Compare the findings

    This paper: Rolling VIX estimate θ̂ = 0.91 ± 0.19 is far below the i.i.d. Monte‑Carlo null (p < 0.001), indicating rough volatility in market data.

    That paper: The Hurst exponent for the range‑based proxies can be as low as 0.014, supporting the rough volatility hypothesis.

    Finding A reports a Hurst exponent as low as 0.014; Finding B reports a rolling VIX estimate θ̂ = 0.91 ± 0.19, both far below null expectations, confirming rough volatility.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.