Correlation Structures and Regime Shifts in Nordic Stock Markets

TL;DR

This study analyzes time-varying correlation structures in Nordic stock markets (OMXS30, OMXC20, OMXH25) to detect regime shifts and improve portfolio robustness. By rolling-window eigenanalysis and Random Matrix Theory-based denoising, it finds that crisis periods exhibit spikes in the leading eigenvalue $\lambda_1$ and counter-cyclic behavior in $\lambda_2$, with the leading eigenportfolio closely tracking the market factor ($\beta_1 \approx 1$, $R^2$ high). The author then develops a regime-aware allocation rule that uses a crisis indicator $\chi(t)=z\{\lambda_1/\lambda_2\}$, cleans the covariance matrix, and constrains exposures to dominant eigenmodes, solving a quadratic program for MV optimization. Backtests show improved downside protection and risk-adjusted performance during stress regimes, while remaining competitive in tranquil periods. Overall, spectral signals offer a practical route to defensive allocation in small, integrated regions like the Nordic markets.

Abstract

Financial markets are complex adaptive systems characterized by collective behavior and abrupt regime shifts, particularly during crises. This paper studies time-varying dependencies in Nordic equity markets and examines whether correlation-eigenstructure dynamics can be exploited for regime-aware portfolio construction. Using two decades of daily data for the OMXS30, OMXC20, and OMXH25 universes, pronounced regime dependence in rolling correlation matrices is documented: crisis episodes are characterized by sharp increases in the leading eigenvalue and counter-cyclical behavior in the second eigenvalue. Eigenportfolio regressions further support a market-factor interpretation of the dominant eigenmode. Building on these findings, an adaptive portfolio allocation framework is proposed, combining correlation-matrix cleaning, an eigenvalue-ratio crisis indicator and long-only minimum-variance optimization with constraints that bound exposures to dominant eigenmodes. Backtesting results indicate improved downside protection and risk-adjusted performance during stress regimes, while remaining competitive with state-of-the-art benchmarks in tranquil periods.

Figures (6)

• Figure 1: Time evolution of the four largest eigenvalues of the correlation matrix of log-returns of the Nordic markets.
• Figure 2: Dynamics of the two largest eigenvalues of the log-return correlation matrix, standardized for comparison sake.
• Figure 3: Dynamics of the largest standardized eigenvalue of the correlation matrix of the Nordic markets.
• Figure 4: Cross-correlation among the largest standardized eigenvalues of the correlation matrices of diverse markets.
• Figure 5: Market regime classification using the crisis indicator $\chi(t)$ and its relation to the dynamics of the standardized leading two eigenvalues for the aggregated Nordic stock market.