Jake XiaChief Risk Officer, Harvard Management Company · Risk Management and Complex Trading

03 Model-Synchronization-Self-Organized-Crowding 2016

PDF · 26 Aug 2026

Share

The paper proposes a unified theoretical framework for understanding synchronization in self‑organized crowds, with a focus on financial market instability but extending to animal herding, bridge sway, traffic jams, and social media. The core is a feedback‑loop model linking three dynamic quantities – agents’ actions (S), the macro observation (O) that agents collectively see, and an external force (E) – through three response coefficients (A, B, C). Agents are assumed to switch between a low‑coupling “normal” state and a high‑coupling “reactive” state when loss‑avoidance thresholds are approached; the proportion of agents in the reactive state (NH/N) determines the aggregate feedback coefficient B. An order parameter R quantifies crowd synchronization as the normalized sum of individual action changes, and is shown analytically to increase with NH/N and decrease with random noise. The model yields a critical tipping point when the product AB reaches unity, beyond which the crowd becomes self‑amplifying, leading to bubbles, crashes, or other unstable phenomena. Simulations illustrate how varying AB and N_H/N generate different market dynamics, including exponential price growth, abrupt crashes, and varying trendiness and volatility. The paper connects the model to existing concepts such as the Vicsek model, Ising model, and ARCH/GARCH volatility models, and argues that modern data sources (social media, smart‑phone apps) could enable measurement of agents’ states, offering new leading indicators of systemic risk. The work emphasizes that while the framework is generic, specific calibration is required for each application, and it complements rather than replaces domain‑specific models.