On microscopic simulation models of financial markets

This thesis makes a number of contributions to the MS literature. First, it develops a Market Fraction (MF) model with heterogeneous raders in a simple asset-pricing framework, which shows that the long-run behaviour and convergence of many variables describing the market can be characterised by the stability and bifurcations of the underlying deterministic system. Next, we characterize various sources of long memory in volatility. Finally, formal econometric techniques are also developed to compare different MS models and evaluate them with respect to market data.