Agent-based and "History-Friendly" Models for Explaining Industrial Evolution

In this paper we compare agent based models (ABMs) and "history-friendly" models (HFMs) of industrial evolution with neoclassical models and discusses their key methodological issues. ABMs are computer-simulated models which incorporate many agents, their actions and interactions with a view to investigating the behavior of a whole system. ABMs are characterized by having bounded rational agents, a learning process, interactions through non-uniform direct networks, disequilibrium, and path-dependency. The strengths and weaknesses of ABMs are addressed in comparison with neoclassical economic models. In particular, ABMs are criticized for their arbitrariness. HFMs are variants of ABMs which aim at capturing in stylized form qualitative theories about mechanisms and factors affecting industrial evolution, technological advances and institutional changes. HFMs consist of three steps: appreciative theories of the history of a specific industry, history-replicating simulations, and history-divergent simulations. In HFMs, model building and calibration are conducted with the guidance of the history. We argue that HFMs overcome the arbitrariness problem in assumptions, and calibration can be justified and falsified by historical evidence. We also discuss some methodological issues related to HFMs and their extendibility. We expect that HFMs will be an important formal modeling tool in evolutionary economics as history-informed economics.