Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

SOE: Fachverband Physik sozio-ökonomischer Systeme

SOE 25: Poster session

SOE 25.16: Poster

Donnerstag, 14. März 2013, 17:15–19:00, Poster C

Dynamic Nonlinearities in Financial Time Series — •Christoph Raeth — Max-Planck-Institut fuer extraterrestrische Physik

The investigation of financial times series by means of nonlinear data analysis is attracting more and more attention in statistical physics. The characteristic fat tails in the probability distribution of the returns are one example of the so-called stylized facts of the fluctuations of price indices. Many of those refer, however, only to the distribution of the returns, where temporal correlations are no longer taken into account.

Here, we present a new method to identify dynamic nonlinearities by analyzing the phase maps of Fourier phases. We find highly significant signatures for nonlinearities in the day-to day returns of the Dow Jones. We repeat the analysis for rank-ordered remapped Dow Jones data, for which by construction the stylized facts referring to fat tails are vanishing. Also in this case we can identify phase correlations. Comparing the data with respective surrogates, we estimate the significance of the detected new signatures of purely dynamic nonlinearities. We further demonstrate that the phase correlations propagate into the calculation of classical nonlinear statistics, e.g. the nonlinear prediction error [1]. We can thus establish qualitatively new signatures in time series of market indices. Realistic market models should be able to reproduce them in addition to the already well-known stylized fact.

[1] C. Raeth, M. Gliozzi, I. E. Papadakis and W. Brinkmann, PRL, 109, 144101 (2012)