[Paper Review] Theories and Practice of Agent based Modeling: Some practical Implications for Economic Planners

Hossein Sabzian, Mohammad Ali Shafia+4 more|arXiv (Cornell University)|Jan 23, 2019

Complex Systems and Time Series Analysis56 references4 citations

TL;DR

This paper provides a systematic, step-by-step framework for developing agent-based models (ABMs) in social sciences, particularly for economic planners, to overcome the limitations of traditional equation-based modeling (EBM) in capturing emergent, complex behaviors. It demonstrates ABM's superiority in modeling adaptive, bounded-rational agents and shows through thought experiments that targeted charity distribution strategies can significantly reduce economic inequality.

ABSTRACT

Nowadays, we are surrounded by a large number of complex phenomena ranging from rumor spreading, social norms formation to rise of new economic trends and disruption of traditional businesses. To deal with such phenomena,Complex Adaptive System (CAS) framework has been found very influential among social scientists,especially economists. As the most powerful methodology of CAS modeling, Agent-based modeling (ABM) has gained a growing application among academicians and practitioners. ABMs show how simple behavioral rules of agents and local interactions among them at micro-scale can generate surprisingly complex patterns at macro-scale. Despite a growing number of ABM publications, those researchers unfamiliar with this methodology have to study a number of works to understand (1) the why and what of ABMs and (2) the ways they are rigorously developed. Therefore, the major focus of this paper is to help social sciences researchers,especially economists get a big picture of ABMs and know how to develop them both systematically and rigorously.

Motivation & Objective

• To address the shortcomings of equation-based modeling (EBM) in capturing emergent, complex behaviors in socio-economic systems.
• To provide social scientists and economic planners with a clear, systematic framework for rigorously developing ABMs.
• To demonstrate the practical utility of ABMs in modeling real-world phenomena such as housing bubbles, inequality, and wealth distribution.
• To highlight the advantages of ABMs over traditional methods like ODEs, PDEs, and system dynamics in modeling adaptive, networked agents.
• To offer a structured development process for ABMs, including modeling, validation, and implementation using tools like NetLogo.

Proposed method

• Adopt a complexity theory and network science foundation to model systems as collections of interacting, adaptive agents.
• Use agent-based modeling (ABM) as a computational method to simulate micro-level rules and local interactions that generate macro-level emergent patterns.
• Implement a bottom-up modeling approach contrasting with top-down EBM, focusing on bounded rationality and adaptability of agents.
• Apply thought experiments using NetLogo to simulate economic systems with heterogeneous agents and charity mechanisms.
• Utilize sensitivity analysis (via BehaviorSpace), participatory simulation (Hubnet), and parameter regime exploration (Behaviorsearch) in NetLogo for model validation.
• Compare different charity allocation strategies (A, B, C) to assess their impact on reducing economic inequality using variance metrics.

Experimental results

Research questions

• RQ1 How can agent-based modeling overcome the limitations of equation-based modeling in capturing emergent behaviors in socio-economic systems?
• RQ2 What are the key methodological steps for systematically and rigorously developing agent-based models in social sciences?
• RQ3 How do different charity distribution strategies affect the reduction of economic inequality in a simulated agent-based economy?
• RQ4 What role does the structure of money transfer (e.g., volume, targeting) play in minimizing variance across economic deciles?
• RQ5 How can ABM frameworks be extended to include richer agent attributes such as memory, education, entrepreneurship, and institutional actors?

Key findings

• ABM outperforms traditional modeling techniques like ODEs, PDEs, and system dynamics in simulating complex, adaptive socio-economic systems with emergent properties.
• The thought experiment shows that charity strategies targeting the poorest agents significantly reduce economic inequality, with strategy C (larger transfers to bottom 50%) showing the greatest decrease in overall average variance.
• The effectiveness of charity in reducing inequality is not only due to higher transfer volumes but also due to the strategic targeting of lower economic deciles.
• Strategy B, despite larger transfers, showed only a marginal improvement in variance reduction compared to strategy C, indicating that targeting matters more than volume alone.
• The simulation results support the Boltzmann-Gibbs law, showing that in a closed economic system, money distribution follows an exponential pattern when conserved.
• Extending ABM to include richer agent features—such as memory, education, and institutional actors—can lead to the emergence of complex behavioral patterns at both micro and macro levels.