Chaotic signals inside some tick-by-tick financial time series.

Abstract

It has been more than four decades since ideas from chaos began appearing in the literature showing that it is possible to design economic models in regime of chaotic behaviour from a theoretical point of view. However there is no clear evidence that economic time series behave chaotically. So far researchers have found substantial evidence for nonlinearity but relatively weak evidence for chaos. In this paper we propose a possible explanation to this ”chaos model-data paradox”. Our main motivation is that chaos is elusive in financial datasets because of loss of information that occurs when daily quotes are used. This could hinder the detection of chaos in those time series. Chaotic systems are sensitive to initial conditions, so temporal dependence is lost as the chaotic time series are sampled at too long-time intervals, appearing as independent even though they come from a (chaotic) dynamical system. In the case of financial time series, which quotes are continuously traded on markets, the daily sampling may be too long. In order to avoid this problem high-frequency data can be used to detect chaos in financial time series. We have found evidence of chaotic signals inside the 14 tick-by-tick time series considered about some top currency pairs from the Foreign Exchange Market (FOREX). Notice that we do not intend to generalize this finding to all financial series or even to all FOREX series. The main interest of our paper is to illustrate that by choosing a tick-by-tick frequency (instead of a daily one), and with the purpose of preserving the dynamic dependence on the time series, we could find chaos. At least in the 14 specific currency pairs analyzed and during the time intervals considered. Hence we propose take into account all the information available in the financial markets (full sample information on FX rates) instead of daily data. This kind of time series entails several difficulties due to the need to process a huge quantity of information and regarding the reconstruction of the attractor from tick-by-tick time series which are unevenly-spaced. In this sense we have had to implemented new algorithms in order to solve such drawbacks. As far as we know these tick-by-tick financial time series have never been tested for chaos so far.