WP7 Stylized complex systems models. Simulations are valuable in and for themselves, but to understand what they mean it is essential that they be reduced to much simpler models, which can be analyzed in mathematical terms. These will be developed in concert with empirical data analysis, using the agent-based models as interpretive filters and for hypothesis testing. The goal will be to get an intuitive understanding of the economy, including nonlinear feedbacks and collective phenomena that are typically not easily addressed in standard economic models based on rational expectations. Such stylized models will not only be useful for researchers, but the intuitive explanations of financial and economic system behaviour they engender will be useful to policymakers as well.

Experience has shown that in parallel with any large-scale ABM effort, it is very helpful to build small-scale, simplified statistical, numerical, and analytical models that help capture, explain, and deepen understanding of core dynamics and behaviours in the large-scale model. Without this, large scale ABMs risk becoming as difficult to explain as the system being modelled itself. The goal of this WP is to explore, using stylized models of the macro-economy how endogenous instabilities and crises emerge at the aggregate level. An important aspect of this research is to identify the important mechanisms leading to such instabilities, both at the level of individual expectation and decision formation, and in terms of interactions and feedback.

To achieve its goals WP7 will:

• study, using analytical methods from statistical mechanics and numerical simulations, stylized agent based models of the macro-economy, in order to understand in particular how and why endogenous instabilities, complex dynamics and crises can emerge at the aggregate level through interactions and feedbacks;

• devise and investigate mathematical models for expectation and decision, alternative to utility maximisation, that take into account limited processing resources, uncertainty, learning & heuristics, framing, habits, etc. and understand the relevance and impact of these effects at the aggregate level.