How to Diversify Differently: Time-Varying Correlations, Determinants, and Regimes
研究了六类资产间15对相关性的时变特征，发现宏观经济变量可预测不同状态，并提出了一个实时状态转换投资策略，相比传统买入持有策略提升了夏普比率等绩效指标。
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.