Network Simulation Engine

The simulation engine provides a framework for evaluating how aggregate liquidity evolves under different network topologies and policy configurations. For each specification, the model generates multiple stochastic paths for the network liquidity multiplier (LM) and identifies episodes in which liquidity falls below a predefined minimum threshold. These episodes are interpreted as liquidity squeeze scenarios. The simulations are used to assess how frequently such events occur, how long they persist, and how their severity responds to alternative policy interventions. In parallel, the framework also identifies states of excess liquidity, which may contribute to asset price inflation.
Network dynamics are modelled through the evolution of the tail exponent, γ, within a preferential attachment framework. Time advances in monthly steps (M), and each network configuration (N) is simulated repeatedly (S) to construct probability distributions of liquidity outcomes. When a sufficiently large adverse topology shock occurs, defined by a threshold change in γ, the network is assumed to remain in a stressed state for a finite number of periods. This mechanism captures the persistence of liquidity stress commonly observed in financial markets.
Changes in γ directly affect the degree distribution of the network. These changes alter the liquidity multiplier and, consequently, the routing capacity of the system. In this way, shocks to network topology translate into endogenous fluctuations in aggregate liquidity, providing a direct link between market structure, liquidity dynamics, and policy effectiveness.
To maintain transparency, the simulation assumes a fixed investment amount per period and a fixed number of network participants. This ensures that liquidity outcomes remain comparable over time and across scenarios. The key inputs are the size of the network (N), the stochastic behavior (lognormal) and persistence of γ, the number of steps (M) and replications (S), transaction volume per period, and the predictive power of the early warning indicator. Policy experiments vary the level of the liquidity buffer and the size and duration of central bank support.
The first is a dynamic liquidity buffer that adjusts when the simulated early warning indicator is triggered. The second is a central bank liquidity injection, modelled as a share of past average network liquidity and distributed uniformly across nodes. Together, these levers allow simulation outcomes to be compared under normal conditions, stressed conditions, and active policy intervention. More targeted interventions, for example those directed at hubs or distressed nodes, are feasible but are left for future extensions.
Model performance is evaluated by comparing distributions of liquidity outcomes across scenarios. The main outputs are the probability that liquidity falls below a squeeze threshold, the duration of such episodes, and the extent to which policy interventions mitigate these risks. In the baseline design, squeeze and surplus thresholds are set at one standard deviation below and above mean network liquidity. This choice enables a transparent comparison of the stabilizing effects of dynamic buffers and central bank support.
Technically, the simulation consists of five building blocks. First, it generates a stochastic path for the topology parameter γ, including persistence following large shocks. Second, each simulated γ is mapped into a network LM. Third, baseline liquidity is computed under a fixed buffer. Fourth, the early warning indicator is allowed to trigger a lower stress buffer and central bank support. Fifth, liquidity outcomes under intervention are calculated.