Anomaly, Class Division, and Decoupling in Wealth Dynamics

Economic inequality is driven by the structure of agent networks, interactions among agents, and the abilities of individual agents. We provide a comprehensive picture of anomalous diffusion, economic class division, and bimodal wealth distribution in an agent-based model, where heterogeneous agent abilities and growth rates are tuned on sparse regular networks. In particular, we focus on the statistical characteristics of logarithmic scaled normalized wealth distributions with two ability parameters, assortativity $\mathcal{A}$ and concentration $\mathcal{R}$. For the set of $(\mathcal{A},\mathcal{R})$, temporal behaviors of log-wealth distributions reveal that the decoupling between different ability groups depends primarily on $\mathcal{R}$ and long-term inequality depends mainly on $\mathcal{A}$. In other words, class division and decoupling are driven by $\mathcal{R}$, while the super-diffusive nature in the leading class is driven by $\mathcal{A}$. Our findings highlight that hierarchical segregation of abilities, rather than ability differences alone, is a key driver of economic class stratification. Finally, our model provides a minimal, yet powerful framework for understanding the bimodal global income distribution observed over the past half century and highlights the critical role of network-level segregation in shaping economic inequality.