Agent-based Models of the Financial System

Jorge A Chan-Lau

Regardless of whether the main data sources are financial statements or security prices, or whether the inter-firm linkages in the financial network correspond to direct exposures or the comovement of security prices, most systemic risk assessment methods and models do not explicitly incorporate the strategic interaction between the agents in the financial system.

Standard bottom-up models analyse the impact of negative shocks in an adverse scenario, where the latter is the outcome of the strategic interactions of the firms and main agents in the financial system. One illustrative example of such a “shock shortcut” modelling approach is the fire sale of assets in balance-sheet network models. Conditional on a scenario where a firm or group of firms have defaulted, the balance-sheet approach assumes that assets must be sold at a loss where the loss calibration uses historical data and/or expert knowledge. The model is silent on why firms end up holding similar asset portfolios or why liquidity providers such as funds and asset management firms fail to enter the market during periods of distress.

The one-off, what-if scenario analysis remains a useful tool for assessing losses