Agent-Based Modelling

A computational approach that simulates many interacting agents and studies the macro patterns that emerge from their behavior.

Agent-based modelling is a way of studying the economy by simulating many individual agents and letting overall outcomes emerge from their interactions. Instead of solving one representative-agent equilibrium directly, the model builds macro behavior from the bottom up.

$$$$

What An Agent-Based Model Contains

An agent-based model usually specifies:

  • agents such as households, firms, banks, or traders
  • rules for how they make decisions
  • an environment in which they interact
  • a timeline for updating behavior, prices, inventories, or beliefs

A simple generic rule is:

[ x_{i,t+1} = f_i(x_{i,t}, s_t, \varepsilon_{i,t}) ]

where x_{i,t} is agent i’s current state, s_t is the surrounding environment, and \varepsilon_{i,t} captures shocks or randomness.

Why Economists Use It

Agent-based modelling is especially useful when heterogeneity matters. A model with different types of households, firms, or banks can generate contagion, clustering, and nonlinear crises that are hard to represent in a single-equation or representative-agent framework.

That is why ABM often appears in work on financial instability, expectations, market microstructure, and complex adaptive systems.

Strengths And Weaknesses

The main strength is flexibility. ABMs can include learning, bounded rationality, network effects, and out-of-equilibrium dynamics.

The main weakness is discipline. Because the model can be very flexible, calibration and validation are harder. Different rule sets can sometimes fit the same aggregate data, so transparency and robustness checks matter a lot.

Where It Fits In Economics

ABM does not replace other modelling traditions. It is best viewed as one tool among many. Standard analytical models are often clearer for identification and welfare results, while ABMs are useful for mechanisms, interactions, and dynamics that are hard to solve analytically.

Knowledge Check

### What makes agent-based modelling different from a representative-agent model? - [x] It builds macro outcomes from the interactions of many individual agents - [ ] It removes heterogeneity from the model entirely - [ ] It assumes every market is always in equilibrium - [ ] It can be used only for tax policy > **Explanation:** ABM works from the bottom up. Many agents interact, and aggregate outcomes emerge from those interactions. ### Why is agent-based modelling attractive for studying crises or contagion? - [ ] Because it guarantees closed-form solutions - [x] Because it can represent heterogeneity, networks, and nonlinear spillovers directly - [ ] Because it ignores shocks - [ ] Because it assumes agents never change behavior > **Explanation:** Crisis dynamics often depend on interaction and heterogeneity, which are natural to represent in ABMs. ### What is one common weakness of agent-based modelling? - [ ] It cannot represent households or firms - [ ] It uses no mathematics at all - [x] It can be difficult to validate and identify because many modeling choices are possible - [ ] It always predicts the same equilibrium as DSGE models > **Explanation:** ABMs can be very flexible, which is useful but can make validation and interpretation harder.