How to Diversify Differently: Time-Varying Correlations, Determinants, and Regimes

This article extends the analysis of the stock–bond correlation to include six diverse asset classes and 15 correlational pairs. The research uncovers significant evidence of recurrent and substantial change in asset correlations. Structural break tests indicate frequent breaks, while wavelet coherence methods highlight widespread temporal instability across both time and frequency domains. Time-varying Granger causality tests reveal significant bidirectional causality between inflation, expected inflation, leading economic indicators, sentiment, and the 15 asset correlations. Logit regressions demonstrate that the same macroeconomic variables predict different regimes based on whether the stock–bond correlations and their returns are rising or falling. We then present a real-time regime-switching portfolio strategy that boosts Sharpe ratios, Sortino ratios, and total returns as well as lowers maximum drawdowns compared to a traditional buy-and-hold benchmark.