ScholarWorks@중앙대학교

초록

Identifying market regimes and modeling their transitions play a pivotal role in understanding market dynamics, risk management, and dynamic portfolio allocation. While Markov-switching models have been widely used to capture these dynamics, they face limitations in scalability and estimation stability as the number of assets or regimes increases. Furthermore, traditional graphical models often rely on static covariance structures, failing to adequately capture the time-varying and non-stationary nature of financial time series. To address these challenges, this paper applies the inverse covariance-based clustering method, adapted from Hallac et al.[40], to analyze structural changes in the Korean stock market using a graph-based approach. The model enables stable estimation in high-dimensional settings and effectively captures temporal shifts in network structures. Specifically, we utilized a dataset spanning five years, consisting of 79 stocks selected from the KOSPI 200 index across the top 30 sectors by market capitalization. This study identified six latent market regimes, analyzed the inter-dependencies among stocks, and conducted a sector-level analysis for each regime. The results demonstrate that the proposed model offers a robust framework for identifying market regimes and understanding the dynamic evolution of the market structure.

키워드