Paper: arXiv 2503.24241

Authors: Hamed Farahani, R. A. Serota

Abstract

We study decades-long historic distributions of accumulated S&P500 returns, from daily returns to those over several weeks. The time series of the returns emphasize major upheavals in the markets – Black Monday, Tech Bubble, Financial Crisis and Covid Pandemic – which are reflected in the tail ends of the distributions. De-trending the overall gain, we concentrate on comparing distributions of gains and losses. Specifically, we compare the tails of the distributions, which are believed to exhibit power-law behavior and possibly contain outliers. Towards this end we find confidence intervals of the linear fits of the tails of the complementary cumulative distribution functions on a log-log scale, as well as conduct a statistical U-test in order to detect outliers. We also study probability density functions of the full distributions of the returns with the emphasis on their asymmetry. The key empirical observations are that the mean of de-trended distributions increases near-linearly with the number of days of accumulation while the overall skew is negative – consistent with the heavier tails of losses – and depends little on the number of days of accumulation. At the same time the variance of the distributions exhibits near-perfect linear dependence on the number of days of accumulation, that is it remains constant if scaled to the latter. Finally, we discuss the theoretical framework for understanding accumulated returns. Our main conclusion is that the current state of theory, which predicts symmetric or near-symmetric distributions of returns cannot explain the aggregate of empirical results.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced statistical methods like linear fits on log-log scales, U-tests for outlier detection, and computes multiple skewness coefficients, indicating moderate to high mathematical complexity. Empirical rigor is strong due to the extensive use of decades-long historical S&P500 data, explicit statistical tests, and analysis of distributions across multiple accumulation periods, making it backtest-ready despite lacking explicit code or portfolio implementation.

Research Flowchart

  flowchart TD
  A["<b>Research Goal</b><br>Explain asymmetry in accumulated<br>S&P 500 return distributions"] --> B["<b>Data Source</b><br>Historic S&P 500 returns<br>Daily to multi-week intervals"]
  B --> C["<b>Methodology</b><br>1. De-trend overall gain<br>2. Analyze tails (power-law)<br>3. U-test for outliers<br>4. PDF & asymmetry analysis"]
  C --> D["<b>Computational Processes</b><br>- Log-log linear fits of CCDF tails<br>- Confidence intervals on fits<br>- Statistical testing<br>- Moment calculations"]
  D --> E["<b>Key Findings</b><br>1. Mean scales near-linearly<br>2. Variance scales perfectly linearly<br>3. Skew remains negative (asymmetric)<br>4. Theory fails to explain results"]

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 to analyze accumulated S&P500 returns, compare the tails of gains and losses using linear fits of complementary cumulative distribution functions on a log‑log scale and a statistical U‑test to detect outliers.

Source sentence
Towards this end we find confidence intervals of the linear fits of the tails of the complementary cumulative distribution functions on a log-log scale, as well as conduct a statistical U-test in order to detect outliers. Namely, we perform linear fits (LF) of the tails of complementary cumulative distribution functions (CCDF) of gains and losses on log-log scale to test for their power-law dependence.

Assumes 3

  • Returns are de‑trended by subtracting a constant daily drift µ, which is assumed to have negligible dependence on the accumulation period τ. model
    Source sentence
    e-trended log returns (or simply ”returns” below) accumulated over time period τ are then given by dxt = xt+τ −xt = rt+τ −rt −µτ = log St+τ St  −µτ (2) where from now on we slide t by one day when obtaining distributions as a function of τ and thus use µ = µ1 – the slope of daily log returns, although clearly µτ shows only very insignificant dependence on τ.
  • Stochastic volatility σ_t follows a mean‑reverting process, implying the mean of the variance vt equals the long‑run level θ. model
    Source sentence
    13 The reason for using mean-reverting models is due to thee assumption that volatility reverts to its mean value, manifested by the fact that the mean of vt is given by vt = θ (11) with the implication that the mean realized volatility for τ-days accumulations is given by V (τ) = Z τ 0 (dxt)2 = Z τ 0 σ2 t dt = Z τ 0 vtdt = θτ (12) Clearly, since σt = √vt is a positive quantity and dW (1) t is even, (7) produces
  • The two Wiener processes driving returns and volatility are uncorrelated, allowing their effects to be treated independently. model
    Source sentence
    Unless one considers such effects as leverage [32, 33] correlations between dW (1) t and dW (2) t can be neglected.

Data

  • S&P500 returns market-prices public
    1980-2024 · S&P500 index
    Source sentence
    We study decades-long historic distributions of accumulated S\&P500 returns, from daily returns to those over several weeks. Fig. 3 shows time series of returns from 1980 to 2024.

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • The mean of de‑trended accumulated returns increases near‑linearly with the number of days of accumulation. statistical
    Source sentence
    The key empirical observations are that the mean of de-trended distributions increases near-linearly with the number of days of accumulation while the overall skew is negative -- consistent with the heavier tails of losses -- and depends little on the number of days of accumulation.
  • The variance of the accumulated return distributions exhibits a near‑perfect linear dependence on the accumulation period, remaining constant when scaled by the number of days. statistical
    Source sentence
    At the same time the variance of the distributions exhibits near-perfect linear dependence on the number of days of accumulation, that is it remains constant if scaled to the latter.
  • The overall skewness of the distributions is negative, indicating heavier tails for losses than for gains, and depends little on the accumulation period. statistical
    Source sentence
    The key empirical observations are that the mean of de-trended distributions increases near-linearly with the number of days of accumulation while the overall skew is negative -- consistent with the heavier tails of losses -- and depends little on the number of days of accumulation. We also found that skewness of the distributions is negative – consistent with the heavier tails of losses – and is only weakly
  • Losses have far heavier power‑law tails than gains, and the power‑law description of the tails deteriorates considerably as the accumulation period increases. statistical
    Source sentence
    The two main findings in this regard is that losses have far heavier tails than gains and that the power-law description of the tails deteriorates considerably with the number of days of accumulation.
  • Current theoretical models that predict symmetric or near‑symmetric return distributions cannot explain the observed empirical asymmetry and mean behavior. theoretical
    Source sentence
    Our main conclusion is that the current state of theory, which predicts symmetric or near-symmetric distributions of returns cannot explain the aggregate of empirical results. which is contrary to what we are seeing empirically.

Checked by others

Supported by

  • Understanding Short-Term Implied Volatility Dynamics: A Model-Independent Approach Beyond Stochastic Volatility (2024)
    On the question: Do short‑term stock return distributions exhibit negative skewness?
    Compare the findings

    This paper: The overall skewness of the distributions is negative, indicating heavier tails for losses than for gains, and depends little on the accumulation period.

    That paper: The estimated log‑return distributions exhibit negative skewness.

    Both findings report negative skewness; B confirms A’s result and shows the effect is stable across accumulation periods.

  • Asymmetric $q$-Gaussian Distributions from Tsallis Maximum Entropy (2026)
    On the question: Do stock return distributions show asymmetry between gains and losses?
    Compare the findings

    This paper: Current theoretical models that predict symmetric or near‑symmetric return distributions cannot explain the observed empirical asymmetry and mean behavior.

    That paper: Anderson–Darling statistics are smaller for the asymmetric model in all 27 market‑horizon combinations.

    B’s asymmetric q‑Gaussian yields lower Anderson–Darling values in all 27 market‑horizon cases, supporting A’s observed asymmetry.

Contradicted by

  • Age and market capitalization drive large price variations of cryptocurrencies (2023)
    On the question: Do positive and negative returns show asymmetric tail behavior in asset price changes?
    Compare the findings

    This paper: The overall skewness of the distributions is negative, indicating heavier tails for losses than for gains, and depends little on the accumulation period.

    That paper: The tail distributions are asymmetric: positive returns more often display smaller exponents, making large positive price variations more likely than negative ones.

    A: positive returns have smaller exponents → larger gains; B: negative skewness → heavier loss tails – they contradict each other.

Cited by 2 papers in the archive

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 (34)

  1. Financial Engineering Playground: Signal Processing, Robust Estimation, Kalman, HMM, Optimization, etc. (2018)
  2. NIST Digital Library of Mathematical Functions, https://dlmf.nist.gov
  3. On the Relation between the Expected Value and the Volatility of the Nominal Excess Return on Stocks (1993)
  4. Distributions of historic market data: relaxation and correlations (2021)
  5. Stochastic volatility and leverage effect (2002)
  6. A theory of the term structure of interest rates (1985)
  7. A model for stock returns and volatility (2014)
  8. A closed-form solution for options with stochastic volatility with applications to bond and currency options (1993)
and 26 more
  1. Probability distribution of returns in the Heston model with stochastic volatility (2002)
  2. Universal behavior of extreme price movements in stock markets (2009)
  3. The distribution of share price changes (1972)
  4. Arch models as diffusion approximations (1990)
  5. Dragon-kings: Mechanisms, statistical methods and empirical evidence (2012)
  6. Are there Dragon Kings in the Stock Market? — → 2307.03693
  7. The Skewness of the Stock Market over Long Horizons (2021)
  8. Skewness in Stock Returns: Reconciling the Evidence on Firm Versus Aggregate Returns (2012)
  9. Good News, Bad News, Volatility, and Betas (1995)
  10. Stock returns and volatility A firm-level analysis (1995)
  11. Asymmetric Volatility and Risk in Equity Markets (2000)
  12. An asymmetric model of changing volatility in stock returns (1992)
  13. The Determinants of Asymmetric Volatility (2001)
  14. Expcted Stock Returns and Volatility (1987)
  15. Estimating and testing skewness in a stochastic volatility model (2023)
  16. Econophysics review: I. Empirical facts (2011)
  17. Empirical properties of asset returns: stylized facts and statistical issues (2001)
  18. Differences of Opinion, Short-Sales Constraints, and Market Crashes (2003)
  19. Rethinking Generalized Beta family of distributions (2023)
  20. Combined Mutiplicative-Heston Model for Stochastic Volatility (2021)
  21. Distributions of Historic Market Data – Stock Returns (2019)
  22. Robust statistical tests of Dragon-Kings beyond power law distribution (2012)
  23. Black swans or dragon-kings? A simple test for deviations from the power law (2012)
  24. Time-scale effects on the gain-loss asymmetry in stock indices (2016)
  25. Temporal structure and gain-loss asymmetry for real and artificial stock indices (2009)
  26. Comparison of Gain–Loss Asymmetry Behavior for Stocks and Indexes (2006)

1 of 34 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.