N°24-77: Dynamic Influence Networks Self-Organize Towards Sub-Critical Financial Instabilities

Instabilities in socio-economic complex systems have long been modeled using Ising-like network interaction models that require fine-tuning to a critical threshold. Recent findings indicate that instabilities can emerge even in sub-critical regimes, provided the network topology is sufficiently non-normal. However, the process by which non-normal networks form is less well understood and has been considered largely decoupled from the dynamics of agents interacting on the network. We show that feedback mechanisms between individual agents and macroscopic quantities such as prices induce feedback loops that cause the network topology to self-organize towards non-normal configurations. The interactions between traders on their dynamically evolving influence network make non-normal networks and financial bubbles intrinsic properties of the financial market dynamics, acting as attractors in the sense of a dynamical system. Through agent-based simulations, we demonstrate that noise traders form a complex network of mutual influences driven by traders’ visibility and success. These dynamics self-organize towards non-normal network topologies that enhance return autocorrelation, increase volatility, and contribute to the formation of financial bubbles, which in turn make the topology more non-normal. We analyze the social trading platform eToro to demonstrate that such feedback mechanisms are active on social trading platforms. Our model thus highlights the increased vulnerability of social systems to instabilities in the presence of global-scale communications.