Paper: arXiv 2504.08611

Authors: Sara A. Safari, Maximilian Janisch, Thomas Lehéricy

Abstract

In the theory of financial markets, a stylized fact is a qualitative summary of a pattern in financial market data that is observed across multiple assets, asset classes and time horizons. In this article, we test a set of eleven stylized facts for financial market data. Our main contribution is to consider a broad range of geographical regions across Asia, continental Europe, and the US over a time period of 150 years, as well as two of the most traded cryptocurrencies, thus providing insights into the robustness and generalizability of commonly known stylized facts.

Complexity vs Empirical Score

  • Math Complexity: 4.5/10
  • Empirical Rigor: 9.0/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper involves moderate mathematical modeling of volatility and statistical testing, but its core focus is on empirical analysis using extensive historical data (150 years) across multiple asset classes and regions, with reproducible code and datasets, making it highly backtest-ready.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Test 11 Stylized Facts"] --> B["Data Collection<br>150 Years: Asia, Europe, US, Crypto"]
  B --> C["Methodology<br>Cross-Asset & Time Horizon Analysis"]
  C --> D["Computational Process<br>Empirical Testing & Validation"]
  D --> E{"Key Outcomes"}
  E --> F["Stylized Facts Robust<br>Across Geographies"]
  E --> G["Time Horizon<br>Dependent Patterns"]
  E --> H["Crypto Markets<br>Exhibit Similar Traits"]

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 systematically evaluates which of eleven stylized facts are supported across diverse markets and time periods, and proposes specific benchmark comparisons to assess intermittency.

Source sentence
Our contribution lies in exploring a larger set of financial data: we use approximately 150 years of traditional financial market data from diverse regions, and also investigate cryptocurrencies, which have different price formation mechanisms than classical financial assets (see [Tan+24; Zha+19]). As standard quantitative tests are lacking, we propose specific comparisons to benchmark processes, and based on these,

Assumes 6

  • The log‑return process is stationary, allowing the autocorrelation function to be estimated by averaging over all time windows. statistical
    Source sentence
    Statistical test Since in practice we only have a single realization of the time series, we estimate the ACF by assuming the stationarity of the process.
  • The statistical tests are model‑agnostic and do not rely on a specific structure of asset returns. model
    Source sentence
    The tests themselves are model-agnostic and we make no assumptions on the specific structure of the asset returns.
  • Residuals of the ACF power‑law fit are approximately normally distributed with constant variance. statistical
    Source sentence
    This estimation assumes that the residuals (the differences between the empirical ACF values [ ACF(l) and the fitted power-law l−β) are approximately normally distributed with a constant variance σ2.
  • Only a single realization of each time series is available, which justifies the use of sample estimators. data
    Source sentence
    Statistical test Since in practice we only have a single realization of the time series, we estimate the ACF by assuming the stationarity of the process.
Show 2 more
  • Cryptocurrency data can be pre‑processed to proxy open/high/low measures despite continuous 24‑hour trading. data
    Source sentence
    As cryptocurrencies operate in a continuous 24-hour market without specific open/high/low conventions, additional pre-processing steps are conducted to proxy these measures.
  • Missing volume values in daily and monthly datasets prevent volume‑volatility analysis for those frequencies. data
    Source sentence
    Result Due to missing values in the traded volume for the daily and monthly datasets, we only report the measurements for the intra-daily datasets in Section 3 for both traditional and cryptocurrency markets.

Data

  • Minute futures data market-prices public
    2012‑2022 · Soybeans, Crude Oil, Nikkei, JPY, FTSE, GBP, EUR‑FX, S&P‑mini
    Source sentence
    Below are the specific datasets used: • Minute data: High-frequency (minute) trading data from futures markets. This data allows the investigation of behaviors in short time frames and includes Soybeans, Crude Oil, Nikkei, Japanese Yen, FTSE, British Pound, Euro FX, S&P-mini from the beginning of 2012 up to the end of 2022. • Daily data: Approximately 10 years of daily futures market data consisting of Canadian
  • Daily futures data · Google Finance market-prices public
    2022‑2023 · Canadian 10‑yr bond, GB gilt, Swiss, Austrian, Korean exchanges, EUR, USD, CNY‑CHF, gold,
    Source sentence
    This data allows the investigation of behaviors in short time frames and includes Soybeans, Crude Oil, Nikkei, Japanese Yen, FTSE, British Pound, Euro FX, S&P-mini from the beginning of 2012 up to the end of 2022. • Daily data: Approximately 10 years of daily futures market data consisting of Canadian 10-year bond, GB long gilt, Switzerland stock exchange, Austrian stock exchange, Korean stock exchange, Euro, USD,
  • Monthly historical data · Global Financial Data / Tick Data market-prices commercial
    1871‑present · Copper, Texas oil, wheat, GB, US, JP 10‑yr bonds, CHF/GBP/JPY‑USD exchange rates
    Source sentence
    This dataset covers traditional market instruments with sufficient granularity to study medium-term phenomena. • Monthly data: Long-term financial data dating back to 1871 for various asset classes and regions. This historical overview enables the examination of stylized facts over extended horizons.
  • Cryptocurrency minute data · Google Finance crypto-onchain public
    2017‑2022 · Bitcoin, Ethereum
    Source sentence
    This dataset consists of copper, Texas oil, wheat, GB, US, and JP 10-year bonds, as well as exchange rates of CHF, GBP, and Japanese Yen to USD. • Cryptocurrency data: minute data for Bitcoin and Ethereum from the beginning of 2017 up to the end of 2022. As cryptocurrencies operate in a continuous 24-hour market without specific open/high/low conventions, additional pre-processing steps are conducted to proxy these

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • Empirical support is found for seven of the eleven stylized facts across the examined markets. statistical
    Source sentence
    Our analysis provides empirical support for seven of the eleven stylized facts.
  • Mixed evidence is observed for the absence of log‑return autocorrelation, differing between traditional markets and cryptocurrencies. statistical
    Source sentence
    We find mixed evidence for the absence of log-return autocorrelation (Section 4.1, differing between traditional markets and cryptocurrencies) and the leverage effect (Section 4.5, not uniformly immediate across all assets).
  • Mixed evidence is observed for the leverage effect, which is not uniformly immediate across all assets. statistical
    Source sentence
    We find mixed evidence for the absence of log-return autocorrelation (Section 4.1, differing between traditional markets and cryptocurrencies) and the leverage effect (Section 4.5, not uniformly immediate across all assets).
  • Results for aggregational Gaussianity suggest convergence toward normality, though this may be driven by fewer tail events at longer horizons. statistical
    Source sentence
    The results for aggregational Gaussianity (Section 4.10) suggest convergence towards normality, but this may be influenced by observing fewer tail events over longer horizons.
  • Evidence supporting intermittency is found, based on benchmark comparisons and visual analysis of volatility distributions. statistical
    Source sentence
    As standard quantitative tests are lacking, we propose specific comparisons to benchmark processes, and based on these, we find evidence supporting intermittency in the data (see Table 3 for a detailed summary). This observation supports the intermittency stylized fact proposed in [Con01].

Checked by others

Qualified by

  • Stylized facts in Web3 (2024)
    On the question: Is a leverage effect present in the asset returns?
    Compare the findings

    This paper: Mixed evidence is observed for the leverage effect, which is not uniformly immediate across all assets.

    That paper: ETH shows a leverage effect, though with lower significance than traditional assets.

    A finds ETH shows a leverage effect with lower significance; B finds mixed evidence and non‑uniform immediacy across assets.

  • Study of Stylized Facts in Stock Market Data (2023)
    On the question: Do log‑return autocorrelations vanish across different financial markets?
    Compare the findings

    This paper: Mixed evidence is observed for the absence of log‑return autocorrelation, differing between traditional markets and cryptocurrencies.

    That paper: Absence of autocorrelations is contradicted in most markets, being confirmed only in Brazil, Canada, and Turkey.

    Both studies report mixed evidence: A finds absence only in Brazil, Canada, Turkey; B finds differing results between traditional markets and crypto.

  • Revisiting Cont's Stylized Facts for Modern Stock Markets (2023)
    On the question: Which stylized facts of financial markets receive empirical support?
    Compare the findings

    This paper: Empirical support is found for seven of the eleven stylized facts across the examined markets.

    That paper: The remaining three stylized facts receive no empirical support in this sample.

    A finds 0 support for the three remaining facts; B finds support for 7 of 11 facts across the examined markets.

Cited by 1 paper in the archive (1 by independent authors)

Code

https://github.com/Saraaqzs/stylized_dev — the authors' own repository

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

Cited References (72)

  1. Empirical properties of asset returns: stylized facts and statistical issues (2001) — link
  2. Scaling in Stock Market Data: Stable Laws and Beyond (1997)
  3. Long range dependence in financial markets (2005)
  4. Volatility Clustering in Financial Markets: Empirical Facts and Agent-Based Models (2007) — link
  5. Scaling and correlation in financial data (1997) — link
  6. Universal features of price formation in financial markets: perspectives from Deep Learning (2018) — link
  7. Econophysics review: I. Empirical facts (2011) — link
  8. Development of an agent-based speculation game for higher reproducibility of financial stylized facts (2019) — link
and 64 more
  1. Revisiting Stylized Facts for Modern Stock Markets (2023) — link
  2. Stylized Facts of High-Frequency Bitcoin Time Series — → 2402.11930
  3. The Self-Organized Criticality Paradigm in Economics & Finance — → 2407.10284
  4. Leverage Effect in Financial Markets: The Retarded Volatility Model (2001) — link
  5. More stylized facts of financial markets: leverage effect and downside correlations (2001) — link
  6. Quadratic Hawkes processes for financial prices (2017) — link
  7. Universality classes for extreme-value statistics (1997) — link
  8. Multivariate Quadratic Hawkes Processes – Part I: Theoretical Analysis (2023) — link
  9. The fine-structure of volatility feedback I: Multi-scale self-reflexivity (2014) — link
  10. Prospect Theory: An Analysis of Decision under Risk (1979)
  11. Volatilities of different time resolutions — Analyzing the dynamics of market components (1997) — link
  12. Heterogeneous volatility cascade in financial markets (2001) — link
  13. ‘Direct’ causal cascade in the stock market from the ‘infrared’ to the ‘ultraviolet’ (1998) — link
  14. A geographical model for the daily and weekly seasonal volatility in the foreign exchange market (1993) — link
  15. Statistical study of foreign exchange rates, empirical evidence of a price change scaling law, and intraday analysis (1990) — link
  16. Multi-scaling in finance (2007) — link
  17. Stylised facts for high-frequency cryptocurrency data (2019) — link
  18. Multiresolution methods for financial time series prediction (1995) — link
  19. From the bird’s eye to the microscope: A survey of new stylized facts of the intra-daily foreign exchange markets (1997) — link
  20. Long-term memories of developed and emerging markets: Using the scaling analysis to characterize their stage of development (2005) — link
  21. SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python (2020) — link
  22. Array programming with NumPy (2020) — link
  23. Data structures for statistical computing in python (2010)
  24. statsmodels: Econometric and statistical modeling with python (2010)
  25. Python 3 Reference Manual (2009)
  26. Time series: theory and methods (1986)
  27. Correlations in economic time series (1997) — link
  28. Agent-based modeling in economics and finance: Past, present, and future (2022)
  29. The Extreme Value Method for Estimating the Variance of the Rate of Return (1980)
  30. Estimating Variance From High, Low and Closing Prices (1991)
  31. Intermittency, Long-Memory and Financial Returns (2007) — link
  32. A MODEL OF PACKET TRAFFIC USING A RANDOM WALL MODEL (1999) — link
  33. Self-Similar Network Traffic and Performance Evaluation (2002) — link
  34. Intermittent transition to turbulence in dissipative dynamical systems. Commun. Math. Phys. 74, 189-197 (1980) — link
  35. A probabilistic approach to intermittency (1999)
  36. Turbulent Cascades in Foreign Exchange Markets (1996) — link
  37. A long memory property of stock market returns and a new model (1993) — link
  38. Critical phenomena in natural sciences: chaos, fractals, selforganization and disorder: concepts and tools (2006)
  39. Introduction to Asset Price Dynamics, Volatility, and Prediction (2007)
  40. A Random-Difference Series for Use in the Analysis of Time Series (1934) — link
  41. The Analysis of Economic Time-Series—Part I: Prices (1953) — link
  42. Every minute counts in financial markets (1991) — link
  43. Intraday periodicity and volatility persistence in financial markets (1997) — link
  44. Dedication (2001) — link
  45. High-Frequency Data and Volatility in Foreign-Exchange Rates (1996) — link
  46. Tests of fit for the logistic distribution based on the empirical distribution function (1979) — link
  47. A closed-form solution for options with stochastic volatility with applications to bond and currency options (1993)
  48. A multifractal model of asset returns (1997)
  49. Intermittency of Superpositions of Ornstein–Uhlenbeck Type Processes (2016) — link
  50. Enriching Financial Datasets with Generative Adversarial Networks (2019)
  51. Generative adversarial networks in finance: an overview (2021) — link
  52. The statistics of bitcoin and cryptocurrencies (2016)
  53. Study of Stylized Facts in Stock Market Data — → 2310.00753
  54. Self-organized criticality: An Explanation of 1/f noise (1987) — link
  55. How nature works (1996) — link
  56. The psycho-biology of language : an introduction to dynamic philology (1935)
  57. Making Sense of Chaos: A Better Economics for a Better World (2024)
  58. Forecasts of Future Prices, Unbiased Markets, and “Martingale” Models (1965) — link
  59. The Levenberg-Marquardt algorithm: Implementation and theory (1978) — link
  60. The effect of intraday periodicity on realized volatility measures (2022) — link
  61. Generalized autoregressive conditional heteroskedasticity (1986) — link
  62. Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation (1982)
  63. Intermittency and nonextensivity in turbulence and financial markets (1999)
  64. Distribution of the Estimators for Autoregressive Time Series With a Unit Root (1979) — link

3 of 72 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.