WP2 Agent-based model of the financial system (FABM). This model will emphasize the network structure of the financial system and allow us to understand and forecast financial instability. The nodes of the model will consist of banks, firms, households, investors, central banks, and regulators. This model will also incorporate realistic agent-behavior and be parameterized with data from WP 1 above. The combination of agent and institutional realism, with explicit network structure of interactions, will enable us to model the formation of bubbles and crises, as well as regimes of stability, from the bottom-up, endogenously.

When it comes to modeling the stability of the interbank market, and of firm-bank credit networks, the available literature reduces to very few contributions that highlight how the interest in these issues., These papers assume a static credit network, and do not take into account the dynamical evolution of the network itself, driven by changing lending/borrowing strategies, at the onset of a credit crisis.

The project will, instead, model agent decision-making explicitly, taking into account the possibility of strategic behaviour and learning as the financial conditions change so as to ensure that the decision-making models for the agents have strong behavioural micro-foundations. This will be achieved by drawing from the output of the experiments run in WP4 and from the on-line games run in WP5.

In order to build the Financial ABM and merge it with the Macroeconomic ABM to be developed in WP3 the following objectives must be reached:

• Extension and harmonization of available ABMs of financial markets. Starting situation: the three groups collaborating in this WP contributed in the past a series of unconnected ABMs and data analysis tools which address different aspects and levels of financial markets. These focus on (1) inter-bank networks; (2) bank-depositors and bank-firm relations; (3) contagion and robustness of banking networks; (4) leverage and regulation. These ABMs have to be extended so that after unification, all systemically relevant issues of financial markets are covered in a single model.

• Understanding systemic risk, originating from inter-bank credit relations, and in particular the relation between contagion, network topology and leverage. Deeper understanding based on existing ABMs is needed regarding the mechanisms of propagation of financial crises to the real economy.

• Unification of various approaches. A common language between the three groups has to be established in terms of programming platforms, data structures and use of variables. A common file server will be installed. Unification of the different ABM components will happen in a modular way. For particular modelling tasks models can be switched on or off in a simulation platform.

• Analysis of empirical data on actual European inter-bank network topologies, bank-firm relations, and data of the real economy (in collaboration with WP1 and WP3) in order to calibrate the model with parameters obtained from real economic data.

• Linking the financial model to the macro model. A central aim is to understand the propagation of financial crashes into the real economy. This will be possible after linking the ABM with those from WP3 into a simulation platform in close collaboration with WP8.

• Policy Implementation. The simulation platform will be user friendly. Documentation addressed to institutions and policy makers containing the major findings will be produced.