Agent Based Models, Housing Fluctuations and the Role of Heterogeneous Expectations

Agent-based models (ABMs) have a long history, but are gradually being introduced as a technique into mainstream economics. ABMs are perhaps best known as a tool for explaining the spatial structure of cities, including patterns of segregation, using cellular automata. In this context ABMs can be used to demonstrate self-organisation and phases of transition, arising from interdependencies in behaviour. A key feature of ABMs is that they relax the traditional assumption of representative agents, used in macroeconomic models, so that agents are heterogeneous in behaviour. This has profound implications for the structure of models of financial and housing markets. In the standard pricing model, where house prices are modelled as the discounted stream of future rental payments, outcomes depend on the choice of the discount rate, which in the case of housing is the user cost of capital. But the user cost of capital requires a measure of house price expectations. Where agents have heterogeneous expectations, the model cannot be solved for the standard rational expectations outcomes and different approaches are required. It cannot be assumed that all agents' expectations are determined by the true model determining prices. The question asked in this paper is what effect on aggregate housing market fluctuations occurs from heterogeneous expectations across agents compared with more conventional representative agent models. The starting point is the Artificial Stock Market Model first developed in the 1990s by Arthur et al., arising from the research programme in complex systems at the Santa Fe Institute, which we adapt to a housing market context and simulate on UK data.