Beyond Computable General Equilibrium: Simulating non-equilibrium dynamics and money, an agent-based Computational Complete Economy

This agent-based model uses the building blocks of a Computable General Equilibrium model, but instead of assuming equilibrium, it explicitly models the price formation and investment processes using money as the medium of exchange. This model gives us several insights that CGE models do not: first, we can check whether or not the results of the CGE models still hold without the equilibrium assumption. Second, as the price formation and investment processes are explicitly modeled, this model reveals non-equilibrium dynamics the evolution of the economy over time. This allows us to conduct experiments that cannot be conducted using CGE models. We can show how the economy reacts to demand or supply shocks and to inflation. In Computational Complete Economy simulations firms act in a physical world. The physical world is derived following the Computable General Equilibrium approach. Based on the social accounting matrix and input-output data, we calibrate the production functions of firms in specific sectors, and use industry data to establish their capacity. Household utility functions are modeled following the CGE approach. On the basis of these physical characteristics, the 30,000,000 firms of the three north American economies are represented as software agents. These agents buy intermediary goods or resources directly from other firms and hire labour from the households. Intermediary goods are sold to companies, and final goods to households. Prices are set by the strategic interaction of firms and by the interaction of firms with households. Thus the modeled economy is a network. Firms can also expand their capacity or abandon establishments. The simulation is based on the ABCE simulation system (AgentBased Complete Economy) a multi-core extension of the SWARM like agent-based modeling protocol for Python. ABCE automatically enforces macroeconomic closedness. The simulation will be open-sourced and can be modified with minimal programming skills.