Refereed journal article or data article (A1)
On the shape of asset return distribution
List of Authors: Töyli Juuso, Kaski Kimmo, Kanto Antti
Publisher: M. Dekker
Publication year: 2002
Journal: Communications in Statistics - Simulation and Computation
Journal name in source: COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION
Journal acronym: COMMUN STAT-SIMUL C
Volume number: 31
Issue number: 4
Start page: 489
End page: 521
Number of pages: 33
ISSN: 0361-0918
eISSN: 1532-4141
DOI: http://dx.doi.org/10.1081/SAC-120004309
URL: https://www.tandfonline.com/doi/full/10.1081/SAC-120004309
Abstract
In this paper we investigate the shape of the asset return distribution using all shares index of Helsinki Stock Exchange and Standard & Poor's 500 index of New York Stock Exchange. In both cases the power exponential distribution is used to model the shape of the return distribution and the inference is cross-checked with Student t-distribution. The possible dependencies in the data are studied by pre-whitening it with GARCH techniques and Cochrane-Orcutt correction. The parameters of the power exponential distribution are estimated with Bayesian approach and with maximum likelihood method. Kolmogorov-Smirnov test, for which the critical values are defined with simulation, is used to test the significance of power exponential fit. The results indicate that there are significant variations in the shape of the distribution over time, which cannot be explained by known time-dependencies. This finding suggests that the shape of distribution might be time-dependent or at least it is non-stationary. In contrast, differences in the shape of the distribution between weekdays are not observed but the tendency towards normality is observed, when the time interval is increased.
In this paper we investigate the shape of the asset return distribution using all shares index of Helsinki Stock Exchange and Standard & Poor's 500 index of New York Stock Exchange. In both cases the power exponential distribution is used to model the shape of the return distribution and the inference is cross-checked with Student t-distribution. The possible dependencies in the data are studied by pre-whitening it with GARCH techniques and Cochrane-Orcutt correction. The parameters of the power exponential distribution are estimated with Bayesian approach and with maximum likelihood method. Kolmogorov-Smirnov test, for which the critical values are defined with simulation, is used to test the significance of power exponential fit. The results indicate that there are significant variations in the shape of the distribution over time, which cannot be explained by known time-dependencies. This finding suggests that the shape of distribution might be time-dependent or at least it is non-stationary. In contrast, differences in the shape of the distribution between weekdays are not observed but the tendency towards normality is observed, when the time interval is increased.
UTU Research Portal