Long-Time Fluctuations in a Dynamical Model of Stock Market Indices

TL;DR

This paper investigates the distribution of stock market index returns using a dynamical model, finding that the central part aligns with Levy distributions while the tails follow a power-law with an exponent greater than 2, reconciling previous empirical discrepancies.

Contribution

It introduces a generic model that captures both Levy-like central fluctuations and power-law tails with exponents above 2, unifying previous empirical findings.

Findings

Central peak scales with Levy distribution

Tails follow a power-law with exponent > 2

Model aligns with empirical data and resolves prior conflicts

Abstract

Financial time series typically exhibit strong fluctuations that cannot be described by a Gaussian distribution. In recent empirical studies of stock market indices it was examined whether the distribution P(r) of returns r(tau) after some time tau can be described by a (truncated) Levy-stable distribution L_{alpha}(r) with some index 0 < alpha <= 2. While the Levy distribution cannot be expressed in a closed form, one can identify its parameters by testing the dependence of the central peak height on tau as well as the power-law decay of the tails. In an earlier study [Mantegna and Stanley, Nature 376, 46 (1995)] it was found that the behavior of the central peak of P(r) for the Standard & Poor 500 index is consistent with the Levy distribution with alpha=1.4. In a more recent study [Gopikrishnan et al., Phys. Rev. E 60, 5305 (1999)] it was found that the tails of P(r) exhibit a power-law decay with an exponent alpha ~= 3, thus deviating from the Levy distribution. In this paper we study the distribution of returns in a generic model that describes the dynamics of stock market indices. For the distributions P(r) generated by this model, we observe that the scaling of the central peak is consistent with a Levy distribution while the tails exhibit a power-law distribution with an exponent alpha > 2, namely beyond the range of Levy-stable distributions. Our results are in agreement with both empirical studies and reconcile the apparent disagreement between their results.