When market boundaries weaken: Network reconfiguration and regime-dependent cross-asset spillovers

Battle Count: 50

Highly relevant for cross-asset portfolio construction, regime-aware risk management, and understanding when diversification benefits break down. The finding that fiat turbulence becomes the dominant transmitter in high-turbulence states and that network clustering/modularity become active transmission channels under stress provides actionable insights for dynamic hedging, crypto-traditional asset allocation, and systemic risk monitoring. However, the paper is primarily diagnostic/descriptive rather than prescriptive for trading strategies.

The framework combines multiple advanced techniques: high-dimensional covariance estimation (Ledoit-Wolf shrinkage), rolling correlation networks with consensus clustering (Louvain + co-membership), Mahalanobis-based turbulence indices, threshold VAR estimation with grid-search threshold selection, Diebold-Yilmaz GFEVD connectedness, and regime-conditional GIRFs with bootstrap inference. Requires careful handling of numerical stability (Moore-Penrose pseudoinverse), stationarity transformations, and multi-step estimation pipelines. The 381-asset system with 9 VAR variables adds computational burden.

The paper provides detailed mathematical formulations for all methods, specifies all hyperparameters (L_net=30, delta_TI=30, lambda=1e-5, theta_rho=0, theta_pi=0.5, B=300 bootstrap replications, H=30 horizon), lists all 381 assets in supplementary tables, and uses publicly available data sources (Yahoo Finance, CBOE, FRED). The code repository is not explicitly provided, but the methodology is fully specified for replication.