Simulation and economic methodology

Abstract

Methodology should reflect on and assess real practices of research, not ideal ones. A practice consists of people working with ideas, tools and techniques to tackle a specific set of problems. This means that beside the more traditional focus on ideas we also must reflect on these tools and techniques. In current practices, the presence of computers is prominent and necessary, and increasingly essential. This even becomes more acute when the computer is becoming more and more a tool of exploration, gradually replacing physical experimentation. In other words, if simulation is becoming the dominant research practice in economics, economic methodologists should pay attention to this. These developments in economic research will change economic methodology. In particular, it will change our conception of explanation and its assessment. Hempel’s nomological-deductive model of explanation will lose this role of being a model for explanation in economics because it’s two constituent elements: laws and deductive logic are too narrow to capture explanations based on simulations. This latter kind of explanations are answers to what-if questions, and not so much, as the covering-law explanations, answers to why-questions. Simulations do not have to be based on realistic models and do not have to be analytically tractable to be rigorous explanations. We only must agree on new rules: rules of construction and rules of validation. This chapter suggests that the first are Turing machines and the second a Turing test.