Regime-Aware Asset Allocation with Dual-Regime Signals and Regime-Dependent Asset Selection

Prior studies in regime-aware asset allocation typically focus on either global marketwide regimes or asset-specific regimes in isolation. However, effective multi-asset allocation requires capturing the interactions and heterogeneity across both marketwide and asset-level dynamics. To address this gap, the authors propose a systems-oriented dual-regime allocation framework with two main contributions. First, it integrates separately estimated global and asset-specific regime signals within a unified allocation framework. Second, it converts interactions between the two regime layers into time-varying, regime-dependent asset universes that adapt to changing markets. The proposed framework consists of three linked components: regime identification, regime forecasting, and regime-dependent portfolio construction. Statistical jump models identify global and asset-specific regimes; XGBoost classifiers forecast next-period regimes; and the joint regime forecasts define two dynamic investable universes: bear market defensive assets (BMDA) and bull market growth assets (BMGA). Empirical results for a multi-asset portfolio from 1990–2025 show that the dual-regime framework outperforms classical regime-switching benchmarks by improving downside resilience and cross-regime diversification through adaptive BMDA/BMGA asset selection.