Complexity and consumption: The role of household behavior in agent-based macro models

Macroeconomic models incorporating a role for heterogeneous-interacting agents (HIA) exhibit fundamentally richer and more realistic dynamics relative to heterogeneous-agent general equilibrium models which ignore interaction. The promise of the agent-based macroeconomic (ABM) approach was to provide more realistic micro-foundations to macroeconomic models by incorporating simply-programmed households, firms, and financial institutions. From the household side, more realistic micro-foundations include behaviorally grounded traits such as bounded rationality, heuristics, and satisficing behavior. These micro features have been shown to create emergent macroeconomic behavior that closely resembles the real world in many aspects. However, ABM modelers have not apparently made much effort to explore varying ways of modeling consumption, focusing instead on labor decisions, research and/or development, and policy. This paper examines the ABM treatment of household consumption, showing that credit and debt play a diminished role in most models. Downplaying the role of household debt shifts the burden of exhibiting macroeconomic patterns to other sectors, where instability may be inappropriately sourced. Issues of stock-flow consistency also cast doubt on the applicability of certain models. Since households do actually save/consume wealth and use credit to smooth consumption, it is likely an important source of instability in the macroeconomy. We make several suggestions to improve agent-based modeling by incorporating richer household behavior.

To date, there is no "standard" ABM model. This heterogeneity has produced two approaches to modeling an economy: incredibly complex simulations that attempt to mimic real economies as closely as possible, and simpler models that attempt to capture "stylized facts" about the economy. Our historical look at agent-based modeling converges in two papers, Caiani et al. (2016), and Lengnick (2013). Both the Caiani paper and the Lengnick papers set out to establish a "baseline'' or "benchmark'' approach to ABM. The history behind these two papers captures a good swathe of the literature that exists. We focus specifically on the development of consumption models through time in relation to agent-based modeling.