Surrogate Data Analysis and Stochastic Chaotic Modelling: Application to Stock Exchange Returns Series

We investigate for evidence of complex-deterministic dynamics in financial returns time series. By combining the Surrogate Data Analysis inferential framework with the MG-GARCH (Kyrtsou and Terraza, 2003) modelling approach, we examine whether the sequences are characterized by aperiodic and nonlinear deterministic cycles or pure randomness. Our results support the hypothesis of complex nonlinear and non-stochastic dynamics in the data generating processes. According to our approach, markets can be assumed to be highly complex, high-dimensional, open and dissipative dynamical systems that need feedback as well as other kinds of inputs in order to operate. These inputs may come in the guise of noise or news. The inputs may also control the evolution of the system dynamics and the knowledge of their nature may allow us to forecast the future states of the market with greater accuracy. To this extent the MG-GARCH model provides a valuable insight on how a feedback mechanism can operate within the structure of stock returns processes and explain stylized facts.