Anticorrelations, subbrownian stochastic drift, and 1/f-like spectra in stable financial systems

Statistic dynamics of financial systems is investigated, basing on a model of randomly coupled equation system driven by stochastic Langevin force. It is found that in stable regime the noise power spectrum of the system is of 1/f^alpha form, with the exponent alpha=3/2 in case of Hermitian coupling matrices, or slightly larger (up to 5/3) in nonhermitian case. In all cases it leads to negative autocorrelation function of the increments of the variables (prices, exchange rates), and to subbrownian stochastic drift of the variables. The model can be generalized to arbitrary stable, driven by noise system of randomly coupled components.