When Factors Collide: Mapping Causal Spillovers across Global Asset Networks
提出CAFNITE模型，将因果推断从模型设定扩展到网络行为，量化全球资产网络中因子间的因果溢出效应，为投资者提供监控因果依赖、压力测试和动态配置因子的实用工具。
Building on the theoretical work of Marcos López de Prado, Alexander Lipton, and Vincent Zoonekynd, who demonstrated that causal factor analysis is necessary for investment efficiency, this article develops a quantitative framework that makes causal factor investing operational. The proposed CAFNITE (<italic>Causal Asset and Factor Network Interference under Treatment Effects</italic>) model extends causal inference from model specification to network behavior, quantifying how shocks to one factor propagate across others through causal linkages. Using a dataset of 20 global factors and 20 assets from 2001 to 2024, CAFNITE estimates direct, indirect, and combined causal effects across equity, bond, currency, and commodity markets. The results reveal asymmetric, state-dependent transmission patterns: during calm regimes, spillovers are diffuse whereas under stress, a small set of factors, particularly the market, momentum, and the US dollar, dominates propagation. By linking causal modeling to factor interaction and diversification diagnostics, CAFNITE provides investors with a practical toolkit for monitoring causal dependencies, stress-testing portfolios, and designing dynamically resilient factor allocations. CAFNITE integrates causal factor inference and interference.