In the last thirty years great advances in theories to explain spatial choice behaviour and in techniques for calibrating spatial choice models have been made. Today there is a broad consensus about what constitutes a state-of-the-art land use model: Except for one (CUFM), which uses allocation rules, all models rely on random utility or discrete choice theory to explain and forecast the behaviour of actors such as investors, households, firms or travellers. Random utility models predict choices between alternatives as a function of attributes of the alternatives, subject to stochastic dispersion constraints that take account of unobserved attributes of the alternatives, differences in taste between the decision makers or uncertainty or lack of information (Domencich and McFadden 1975). Anas (1983) showed that the multinomial logit model resulting from random utility maximisation is, at equal levels of aggregation, formally equivalent to the entropy-maximising model proposed by Wilson (1967, 1970); he thus laid the foundation for the convergence and general acceptance of formerly separate strands of theory. The STASA model is based on the master equation approach and may be seen as a dynamic and decision based multi-agent system (Haag, 1990). Underneath that uniformity, however, there are significant differences between the theoretical foundations of the models: