Agent-based modelling
Explore aggregate patterns generated by interacting agents
Agent-based modelling represents interacting entities with explicit decision rules, information and adaptation, then examines the aggregate patterns their interactions generate. Financial models may include traders, banks or networks with heterogeneous behaviours. Unlike a reduced-form forecasting equation, the model proposes a generative mechanism; many alternative rule sets can still produce similar aggregate patterns, making validation and comparison essential.
Choose this when heterogeneity, interaction or adaptation is central and aggregate equations obscure the mechanism of interest. Agent rules must be explainable and the emergent outcomes compared with meaningful empirical or theoretical benchmarks.
Strengths
- Represents heterogeneous behaviour and feedback explicitly
- Allows controlled examination of mechanisms that are hard to isolate empirically
Limitations
- Many mechanisms can reproduce the same stylised facts
- Calibration, computation and sensitivity analysis can be demanding
Know the boundary
Reproducing volatility clustering or another stylised fact does not uniquely identify the real behavioural mechanism.