In this paper we want to discuss macroscopic and microscopic properties of financial markets. By analyzing quantitatively a database consisting of 13 minute per minute recorded financial time series, we identify some macroscopic statistical properties of the corresponding markets, with a special emphasize on temporal correlations. These analysis are performed by using both linear and nonlinear tools. Multivariate correlations are also tested for, which leads to the identification of a global coupling mechanism between the considered stock markets. The application of a new formalism, called transfer entropy, allows to measure the information flow between some financial time series. We then discuss some key aspects of recent attemps to model financial markets from a microscopic point of view. One model, that is based on the simulation of the order book, is described more in detail, and the results of its practical implementation are presented. We finally address some general aspects of forecasting and modeling, in particular the role of stochastic and nonlinear deterministic processes.
time series analysis econophysics simulated markets temporal correlations high-frequency data