Paper: arXiv 2409.10543

Authors: L. Ponta, A. Carbone

Abstract

The Kullback-Leibler cluster entropy $\mathcal{“D_{C”}}[“P | Q”] $ is evaluated for the empirical and model probability distributions \(P\) and \(Q\) of the clusters formed in the realized volatility time series of five assets (SP&500, NASDAQ, DJIA, DAX, FTSEMIB). The Kullback-Leibler functional $\mathcal{“D_{C”}}[“P | Q”] $ provides complementary perspectives about the stochastic volatility process compared to the Shannon functional \(\mathcal{“S_{C”}}[“P”]\). While $\mathcal{“D_{C”}}[“P | Q”] $ is maximum at the short time scales, \(\mathcal{“S_{C”}}[“P”]\) is maximum at the large time scales leading to complementary optimization criteria tracing back respectively to the maximum and minimum relative entropy evolution principles. The realized volatility is modelled as a time-dependent fractional stochastic process characterized by power-law decaying distributions with positive correlation (\(H>1/2\)). As a case study, a multiperiod portfolio built on diversity indexes derived from the Kullback-Leibler entropy measure of the realized volatility. The portfolio is robust and exhibits better performances over the horizon periods. A comparison with the portfolio built either according to the uniform distribution or in the framework of the Markowitz theory is also reported.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.0/10 (capped from 7.0: transaction costs not modelled; no out-of-sample test; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 7.0)

Why this score: The paper employs advanced mathematical concepts like Kullback-Leibler divergence, coarse-graining, and fractional stochastic processes with Hurst exponents, indicating high math complexity. It also demonstrates strong empirical rigor by applying the methodology to real financial data (five major assets), constructing a portfolio based on the results, and comparing performance against established benchmarks like Markowitz.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Quantify Volatility Correlation &<br>Risk Diversity using KL Divergence"] --> B["Data Collection:<br>Realized Volatility of 5 Assets<br>SP500, NASDAQ, DJIA, DAX, FTSEMIB"]
  B --> C["Methodology:<br>Cluster Time-Series &<br>Calculate KL Cluster Entropy<br>vs Shannon Entropy"]
  C --> D["Modeling:<br>Fractional Stochastic Process<br>Power-law distributions H>1/2"]
  C --> E["Portfolio Optimization:<br>Diversity Index derived from<br>KL Entropy Measure"]
  D --> F["Key Findings:<br>1. KL max at short scales<br>2. Shannon max at large scales<br>3. KL-based portfolio<br>outperforms Markowitz"]
  E --> F

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 proposes using Kullback‑Leibler cluster entropy to create diversity measures for realized volatility and to build a robust multi‑period portfolio based on these measures.

Source sentence
The Kullback-Leibler cluster entropy will be used to define a robust and sound set of diversity measures built upon the coarse grained probability distribution of the realized volatility of tick‑by‑tick data of the Standard & Poor 500 (SP&500), Dow Jones Ind. Avg (DJIA), Deutscher Aktienindex (DAX), Nasdaq Composite (N

Assumes 5

  • The model probability distribution Q is assumed to follow a power‑law rather than a Gaussian distribution. model
    Source sentence
    The cluster relative entropy described in Section IIA is conditioned by the model probability Q, which is assumed to be a power law rather than on an unrealistic hypothesis of Gaussianity of returns.
  • The realized volatility series are made stationary by a detrending procedure. data
    Source sentence
    The functionals D(τj, n) are estimated on the realized volatility series of the financial market data, made stationary by the detrending procedure.
  • The underlying volatility process is a time‑dependent fractional stochastic process with Hurst exponent H > 1/2. model
    Source sentence
    The realized volatility is modelled as a time‑dependent fractional stochastic process characterized by power‑law decaying distributions with positive correlation (H>1/2).
  • Weights wi,D are estimated from Eq. (15) applied to the realized volatility of the five assets. statistical
    Source sentence
    Next, the weights wi,D as defined by Eq. (15) are estimated for the realized volatility of the five assets.
Show 1 more
  • Investors can rebalance monthly without market impact, using adjusted close prices as given. execution
    Source sentence
    The number of shares of each index is updated every month evaluating the initial wealth allocated to each index (dividing Wo by the weight of the corresponding month) dividing by the price of each index of the corresponding month.

Data

Backtest sample: 2018-01 – 2018-12 · S&P 500, NASDAQ, DJIA, DAX, FTSEMIB indexes · equities · monthly

  • Tick‑by‑tick realized volatility data market-prices public
    2018 (twelve monthly periods) · SP&500, NASDAQ, DJIA, DAX, FTSEMIB
    Source sentence
    The Kullback-Leibler cluster entropy and the diversity indexes are estimated over twelve monthly periods covering the year 2018.

Costs

Not modeled A trading or execution result is reported without costs.

Finds 2

  • The Kullback‑Leibler‑based portfolio outperforms both the equally‑weighted and Markowitz (Sharpe‑ratio) portfolios for active and lazy investor strategies. performance
    Source sentence
    It is worth noting that the Kullback‑Leibler weigths portfolio shows the best performances both with the active and lazy investor strategy compared to the equally weighted
  • Kullback‑Leibler weights wi,D vary smoothly with investment horizon and tend toward uniform weights for longer volatility windows. statistical
    Source sentence
    The relative cluster entropy weights wi,D vary continuously and take values close to the equally distributed weights ui. At short investment horizons M, the weights wi,D deviate more from the equally distributed values than they do at long investment horizons.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 9 related papers). Coverage is partial, so this is neither replication nor refutation.

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.

Cited References (43)

  1. Learning from Biodiversity: Is Diversity in Financial Ecosystems Important for Economic Growth and Stability? (2024) — link
  2. Relationships between tail entropies and local intrinsic dimensionality and their use for estimation and feature representation (2023) — link
  3. Performance-guaranteed regularization in maximum likelihood method: Gauge symmetry in Kullback-Leibler divergence. (2023) — link
  4. Analysis of temporal structure of laser chaos by Allan variance. (2022) — link
  5. Relative cluster entropy for power-law correlated sequences (2022) — link
  6. Learning Uncertainties the Frequentist Way: Calibration and Correlation in High Energy Physics. (2022) — link
  7. Rough Volatility: Fact or Artefact? (2022) — link
  8. Inferring multi-period optimal portfolios via detrending moving average cluster entropy (a) Contribution to the Focus Issue Progress on Statistical Physics and Complexity edited by Roberta Citro, Giorgio Kaniadakis, Claudio Guarcello, Antonio Maria Scarfone and Davide Valenti. (2021) — link
and 35 more
  1. The maximum entropy of a metric space (2019) — link
  2. Information measure for financial time series: Quantifying short-term market heterogeneity (2017) — link
  3. What is the Consumption-CAPM missing? An information-theoretic framework for the analysis of asset pricing models (2017) — link
  4. Entropy-Based Financial Asset Pricing (2014) — link
  5. Volatility is rough (2014) — link
  6. Clustering by fast search and find of density peaks (2014) — link
  7. Information Measure for Long-Range Correlated Sequences: the Case of the 24 Human Chromosomes (2013) — link
  8. Kullback–Leibler quantum divergence as an indicator of quantum chaos (2012) — link
  9. Entropy production and Kullback-Leibler divergence between stationary trajectories of discrete systems. (2012) — link
  10. Between order and chaos (2011) — link
  11. Sources of Entropy in Representative Agent Models (2011) — link
  12. Information Theoretic Measures for Clusterings Comparison: Variants, Properties, Normalization and Correction for Chance (2010) — link
  13. Data clustering: 50 years beyond K-means (2008) — link
  14. Information Theory, Relative Entropy and Statistics (2008) — link
  15. Optimal Portfolio Diversification Using the Maximum Entropy Principle (2008) — link
  16. Scaling properties and entropy of long-range correlated time series (2007) — link
  17. Kullback-Leibler distance as a measure of the information filtered from multivariate data. (2007) — link
  18. Numerical Methods in Finance and Economics: A MATLAB-Based Introduction (2006) — link
  19. Analysis of clusters formed by the moving average of a long-range correlated time series. (2004) — link
  20. A Guide to First‐passage Processes (2003) — link
  21. MODELING AND FORECASTING REALIZED VOLATILITY (2003) — link
  22. Econometric analysis of realized volatility and its use in estimating stochastic volatility models (2002) — link
  23. Information Theoretic Clustering (2002) — link
  24. The role of relative entropy in quantum information theory (2001) — link
  25. Long memory in continuous‐time stochastic volatility models (1998) — link
  26. Distribution of the first return time in fractional Brownian motion and its application to the study of on-off intermittency. (1995) — link
  27. The dynamic origin of increasing entropy (1989) — link
  28. Characterization of Strange Attractors (1983) — link
  29. Diversity and Evenness: A Unifying Notation and Its Consequences (1973) — link
  30. MULTI-PERIOD MEAN-VARIANCE ANALYSIS: TOWARD A GENERAL THEORY OF PORTFOLIO CHOICE* (1971) — link
  31. Portfolio Selection (1971) — link
  32. Kowalewska-Kud(cid:32)laszyk (2012)
  33. Model selection and multi-model inference. Second. (2004)
  34. Markowitz Revisited: Mean-Variance Models in Financial Portfolio Analysis (2001) — link
  35. active investor regularly updating the portfolio shares (monthly in our example)

0 of 43 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.