Essays on Aggregation With Heterogeneous and Interacting Agents

Abstract

Aggregation is one, if not the fundamental challenge of the social sciences. Whenever agents on the micro level interact and are strongly heterogeneous, macro behavior might exhibit emergent properties and is likely irreducible to an ensemble of isolated actions at the micro level. Given the overwhelming complexity of those actions with myriads of agents interacting based on differing beliefs and along heterogeneous dimensions, any explanatory attempt needs to idealize and simplify to reduce this complexity. While the focus of behavioral macroeconomics is to refine behavioral rules, the interaction and heterogeneity of those sophisticated agents can therefore often only be considered to a limited extent. By contrast, the complexity economics approach typically assumes little sophistication and adaptive behavior and focuses on the impact of the intermediate layer between the micro and macro levels on aggregate outcomes, i.e., on distributional regularities and networks. This is the paradigm in which the thesis at hand is situated. Which approach is more appropriate to understand a specific economic phenomenon naturally depends on the network topology and actual heterogeneity of the empirical target system. Rather than substitutes, the central claim this thesis sets out to defend is thus that both perspectives complement each other and that a complexity perspective offers additional interpretations and implications concerning the same phenomenal domain. Instead of a conceptual discussion, it tries to employ the outlined pre-analytic vision and shows that the proposed parsimonious explanations are often observationally equivalent to attempts that attribute most of the observed macro outcomes to micro origins. In these cases, a complexity economics perspective might offer vastly different additional policy implications, e.g., it might help to identify pivotal agents for aggregate outcomes or show how changing social network topologies might influence mental models that more micro-oriented approaches cannot account for by construction.