High-Dimensional Industry Network Based on Adaptive T-Lasso Algorithm

This paper investigates the volatility connectivity among China's industry ETFs, constructing a comprehensive high-dimensional network that includes 96 industry ETFs. Given the complexity and dynamic nature of financial markets, we utilize the vector autoregression (VAR) model as the foundational analytical tool. However, recognizing the challenges posed by high-dimensional data and the need to account for the fat-tailed distribution characteristic of financial data, we introduce the Adaptive t-lasso method. This method improves prediction accuracy by effectively addressing consistency issues in high-dimensional variable selection. Our empirical analysis identifies significant influences from key industries such as electronics, real estate, non-bank financials, and defense while highlighting strong interconnections in sectors like information technology, new energy, and semiconductors. These findings illuminate the intricate linkages between industries across various market cycles. The results provide not only a more robust framework for analyzing industry volatility but also essential insights for investors navigating the complexities of a market environment.