Sample period selection and long-term dependence: New evidence from the Dow Jones index

Jonathan A. Batten*, Craig A. Ellis, Thomas A. Fethertson

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

15 Citations (Scopus)

Abstract

This study employs the classical and modified rescaled adjusted range statistic (R/S statistic) to investigate the sensitivity of the long-term return anomaly observed on the Dow Jones Industrial Average (DJIA) to sample and method bias. Daily data from 1/1/1970 to 17/3/2004 is used with sub-periods identified based on sign shifts in the mean returns as well as the October 1987 crash. The return series are also filtered to accommodate autoregressive conditional heteroskedastic (ARCH) innovations and short-term dependencies. Hurst exponent and V-statistic values for each of the filtered series for the whole sample and sub-periods are estimated, while polynomial regression techniques are applied to plot the V-statistics. These plots show oscillating cycles of varying lengths. Overall, we find the null hypothesis of no long-term dependence is accepted for the whole sample and every sub-period using the modified rescaled range test, but not necessarily using the classical rescaled adjusted range test. The later test does, however, reveal episodes of both positive and negative dependence over the various sample periods, which have been reported by other researchers.

Original languageEnglish
Pages (from-to)1126-1140
Number of pages15
JournalChaos, Solitons and Fractals
Volume36
Issue number5
DOIs
Publication statusPublished - Jun 2008

Fingerprint

Dive into the research topics of 'Sample period selection and long-term dependence: New evidence from the Dow Jones index'. Together they form a unique fingerprint.

Cite this