Empirical Models, Rules, And Optimization: Turning Positive Economics On Its Head

This paper considers supply decisions by firms in a dynamic setting with adjustment costs and compares the behavior of an optimal control model to that of a rule-based system which relaxes the assumption that agents are explicit optimizers. In our approach, the economic agent uses believably simple rules in coping with complex situations. We estimate rules using an artificially generated sample obtained by running repeated simulations of a dynamic optimal control model of a firm’s hiring/firing decisions. We show that (i) agents using heuristics can behave as if they were seeking rationally to maximize their dynamic returns; (ii) the approach requires fewer behavioral assumptions relative to dynamic optimization and the assumptions made are based on economically intuitive theoretical results linking rule adoption to uncertainty; (iii) the approach delineates the domain of applicability of maximization hypotheses and describes the behavior of agents in situations of economic disequilibrium. The approach adopted uses concepts from fuzzy control theory. An agent, instead of optimizing, follows Fuzzy Associative Memory (FAM) rules which, given input and output data, can be estimated and used to approximate any non-linear dynamic process. Empirical results indicate that the fuzzy rule-based system performs extremely well in approximating optimal dynamic behavior in situations with limited noise. Simulations are also performed under increasingly noisy.