Industry Portfolio Volatility Connections and Industry Portfolio Returns
使用贝叶斯时变参数VAR模型追踪美国行业投资组合收益波动率的动态关联，发现这些波动率关联形成的网络结构能预测未来行业投资组合收益，对点预测和密度预测均有显著提升。
ABSTRACT This paper tracks dynamic connections that form among daily US industry portfolio return volatilities using a Bayesian time‐varying parameter VAR model. Market participants often focus on sectors to filter vast amounts of information, and this focus results in cross‐industry return predictability. We characterise connections that form over the short‐, medium‐ and long‐term, analysing their role as indicators of sectoral uncertainties. These volatility‐based connections create a network structure that reflects co‐movement across different industries. By capturing these network dynamics, we assess their usefulness in predicting US industry portfolio returns. Our results show that network connections contain economically meaningful information for future industry portfolio returns, offering significant gains in both point and density forecasts.