COMPUTATIONAL SOCIAL SCIENCE

Department of Computational Social Science Seminar Abstract

Friday, December 3 - 3:00 p.m.

Endogenous Macroeconomic Fluctuations from Purely Microeconomic Phenomena in a Model with 150 Million Agents

Robert Axtell, Chair
Department of Computational Social Science
George Mason University

ABSTRACT: Conventional macroeconomic models have microeconomic specifications in which the agents (workers, firms, consumers) who make up the economy are assumed to be in general equilibrium. As a result, the only sources of dynamics in such model economies are external shocks. These shocks are treated as stochastic, ostensibly outside the bounds of normal economic theorizing. However, a wide-range of economic downturns and financial crises appear to have no external cause, but rather seem to be due to processes internal to their underlying economies. Here we show that a microeconomic model, written at the level of individual economic agents, can endogenously produce macroeconomic dynamics that closely resemble those of modern macroeconomies: idiosyncratic business cycles, waves of involuntary unemployment, irregular exponential growth, spontaneous firm birth and death, job turnover, skew income distributions, fluctuating prices, power law distributions of firm sizes, heavy-tailed firm growth rates, and inflation. The model is built from millions of software agents and realized computationally. Agents work in team production environments, are paid wages, migrate between firms based on job opportunities, and start new firms when they perceive it advantageous to do so. Prices are local and fluctuate endogenously. Consumers seek utility by purchasing goods from firms. Firms come and go based on sales and profitability. Microeconomic equilibria exist but are dynamically unstable. No macroeconomic specifications of any kind are employed in the model. Overall, aggregate structure emerges endogenously from the bottom up, through the interactions of the purposive agents. Specifically, the aggregate economy expands with fluctuations about a steady-state growth rate while the agent level is always out of equilibrium. As the economy grows the fluctuations decay in a peculiar way that is consistent with empirical data.