Paper: arXiv 2509.19663
Authors: Yifan He, Svetlozar Rachev
Abstract
This study presents a comprehensive empirical investigation of the presence of long-range dependence (LRD) in the dynamics of major U.S. stock market indexes–S&P 500, Dow Jones, and Nasdaq–at daily, weekly, and monthly frequencies. We employ three distinct methods: the classical rescaled range (R/S) analysis, the more robust detrended fluctuation analysis (DFA), and a sophisticated ARFIMA–FIGARCH model with Student’s \(t\)-distributed innovations. Our results confirm the presence of LRD, primarily driven by long memory in volatility rather than in the mean returns. Building on these findings, we explore the capability of a modern deep learning approach, Quant generative adversarial networks (GANs), to learn and replicate the LRD observed in the empirical data. While Quant GANs effectively capture heavy-tailed distributions and some aspects of volatility clustering, they suffer from significant limitations in reproducing the LRD, particularly at higher frequencies. This work highlights the challenges and opportunities in using data-driven models for generating realistic financial time series that preserve complex temporal dependencies.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 6.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced statistical and econometric methods like ARFIMA–FIGARCH models and deep learning (Quant GANs), indicating high mathematical density, while its reliance on real-world financial data and backtesting-like simulation analysis shows strong empirical components.
Research Flowchart
flowchart TD
A["Research Goal:<br>Investigate LRD in U.S. Stock Markets"] --> B["Data Collection:<br>S&P 500, DJI, NASDAQ"]
B --> C["Methodology:<br>R/S, DFA, ARFIMA-FIGARCH"]
C --> D["Process: <br>Empirical Analysis of Returns & Volatility"]
D --> E{"Finding 1:<br>Presence of LRD?"}
E -->|"Yes"| F["Outcome:<br>LRD driven by Volatility"]
E -->|"No"| G["Outcome:<br>No Significant LRD"]
F --> H["Generative Modeling:<br>Quant GANs"]
G --> H
H --> I{"Outcome 2:<br>Can GANs replicate LRD?"}
I -->|"Limited"| J["Finding:<br>GANs fail to capture LRD,<br>especially at high frequencies"]
I -->|"Yes"| K["Finding:<br>GANs successfully replicate LRD"]
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 conducts a comprehensive empirical investigation of long-range dependence in major U.S. stock indexes and explores the use of Quant generative adversarial networks to learn and replicate this dependence.
Source sentence
This study presents a comprehensive empirical investigation of the presence of long-range dependence (LRD) in the dynamics of major U.S. stock market indexes--S\&P 500, Dow Jones, and Nasdaq--at daily, weekly, and monthly frequencies. Building on these findings, we explore the capability of a modern deep learning approach, Quant generative adversarial networks (GANs), to learn and replicate the LRD observed in the
Assumes
No explicit assumptions found in the text.
Data
- Daily closing values of S&P 500, Dow Jones, and Nasdaq market-prices public
Source sentence
2.1 Three Benchmark Indexes of the U.S. Stock Market The original dataset used in this study consists of the daily closing values of three benchmark indexes of the U.S. stock market–S&P 500, Dow Jones, and Nasdaq–covering the period from January 2, 1992, to December 30, 2024, for a total of 8310 trading days, retrieved using the yfinance1 package in Python, with Yahoo Finance2 serving as the data source.
Costs
Not applicable
Finds 5
- Long-range dependence is present in the index series, primarily driven by long memory in volatility rather than in mean returns. statistical
Source sentence
Our results confirm the presence of LRD, primarily driven by long memory in volatility rather than in the mean returns.
- Quant GANs effectively capture heavy‑tailed return distributions but suffer significant limitations in reproducing LRD, especially at higher frequencies. negative
Source sentence
While Quant GANs effectively capture heavy-tailed distributions and some aspects of volatility clustering, they suffer from significant limitations in reproducing the LRD, particularly at higher frequencies.
- Detrended fluctuation analysis shows that the apparent LRD in the Dow Jones index is likely spurious and attributable to short‑term correlations. statistical
Source sentence
However, it is evident that the Dow Jones index, across all three frequencies, does not exhibit significant LRD, whereas the S&P 500 and Nasdaq indices do.
- ARFIMA‑FIGARCH modeling indicates that the observed LRD is predominantly a feature of conditional volatility (dv ≈ 0.4) rather than the conditional mean of returns. statistical
Source sentence
The ARFIMA-FIGARCH modeling provided the most nuanced insight, clearly indicating that the observed LRD is predominantly a feature of conditional volatility (captured by dv ≈0.4) rather than the conditional mean of returns.
- Synthetic data generated by Quant GANs largely fails to replicate the LRD inherent in the original empirical series, particularly at daily and weekly frequencies. negative
Source sentence
However, a rigorous evaluation using the same battery of LRD tests revealed a significant shortcoming: the synthetic data generated by Quant GANs largely failed to replicate the LRD inherent in the original empirical series, particularly at daily and weekly frequencies.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 23 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 (32)
- An empirical evaluation of generic convolutional and recurrent networks for sequence modeling (2018) — link
- Fractionally integrated generalized autoregressive conditional heteroskedasticity (1996)
- On H urst exponent estimation under heavy-tailed distributions (2010)
- Statistics for Long-Memory Processes (2017)
- A conditionally heteroskedastic time series model for speculative prices and rates of return (1987)
- Multinational Finance: Evaluating the Opportunities, Costs, and Risks of Multinational Operations (2016)
- Alternative Investments: CAIA Level I (2015)
- Empirical properties of asset returns: Stylized facts and statistical issues (2001)
and 24 more
- Theory and Applications of Long-Range Dependence (2002)
- A suggestion for using powerful and informative tests of normality (1990)
- Value-at-Risk and extreme returns (2000)
- Modelling Extremal Events: For Insurance and Finance (2013)
- The behavior of stock-market prices (1965)
- Generative adversarial nets (2014)
- TCGAN : Convolutional Generative Adversarial Network for time series classification and clustering (2023)
- PSA-GAN : Progressive self attention GAN s for synthetic time series (2021)
- Sig-Wasserstein GANs for time series generation (2021)
- COT-GAN : Generating sequential data via causal optimal transport (2020)
- Time-series generative adversarial networks (2019)
- An introduction to long-memory time series models and fractional differencing (1980)
- Long-term storage capacity of reservoirs (1951)
- Detecting long-range correlations with detrended fluctuation analysis (2001)
- The jackknife and the bootstrap for general stationary observations (1989)
- Long-term memory in stock market prices (1991)
- The variation of certain speculative prices (1963)
- Noah, J oseph, and operational hydrology (1968)
- Mosaic organization of DNA nucleotides (1994)
- An Introduction to Analysis of Financial Data with R (2014)
- Optimal Transport: Old and New (2009)
- Estimating long-range dependence: Finite sample properties and confidence intervals (2002)
- Skewed multifractal scaling of stock markets during the COVID-19 pandemic (2023)
- Q uant GAN s: Deep generation of financial time series (2020)
0 of 32 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.