Cellular Genetic Automata in Computer Simulation of Economic Growth and Development with Romer Externalities

Roger A. McCain - Drexel University

The present study investigates growth and economic development in an overlapping generations population of two-period lived agents situated on a cellular grid. Each cell in each period is occupied by one newly born young agent and one old agent. Income in each cell is determined as a function of the physical capital resulting from the old agent's investment when young, the labor and human capital supplied by the young agent, and the average human capital supplied in neighboring cells (the "Romer externality"). Each young agent must reduce work time in order to study and form human capital and must consume less than wage income in order to form physical capital for the following period. Each old agent consumes the profit or interest share of the income produced in his cell. In the initial simulation period, each young agent's decisions concerning human and physical capital formation are randomly determined. In subsequent periods these decisions are evolved by means of a genetic algorithm using one of two possible specifications: global learning, in which genetic operations are applied to the agent population as a whole; and localized learning in which genetic operations are applied only to neighborhoods. The variation of production over the cellular automaton grid is visualized as a surface and displayed for various degrees of localized learning and localized Romer externalities. Particular attention is paid to systematic patterns in the formation and spread of rich and poor regions under these various specifications.

Scheduled for Session 2.6 Agent - Based Computational Economics - I