High uncertainty and increasing interconnectivity in global financial markets challenge static asset allocation strategies, particularly in emerging economies. This study evaluates the effectiveness of three volatility proxies (conditional, implied, and historical) in identifying market regimes through a Hidden Markov Model and examines their economic relevance for ETF allocation in Brazil and the United States using daily data from 2018 to 2024, with portfolio evaluation focused on the 2020–2024 out-of-sample period. The results indicate that conditional volatility estimated via GARCH provides more stable and operationally reliable regime identification. Model selection based on the Bayesian Information Criterion supports a three-regime specification (low, medium, and high volatility), allowing the model to capture not only extreme market states but also transitional phases often overlooked in binary frameworks. A key contribution of the study is the identification of an international risk transmission mechanism. Granger causality tests applied to regime probabilities reveal that increases in the probability of high-volatility regimes in the U.S. market precede regime shifts in the Brazilian market, suggesting that U.S. volatility regimes function as an early warning system for emerging markets. In asset allocation, the dynamic regime-based strategy outperforms static mean–variance optimization (Single Regime) in risk-adjusted terms in Brazil while successfully mitigating severe losses in the U.S. market. However, the dynamic strategy does not outperform the naive equal-weighted (1/N) benchmark, highlighting the trade-off between dynamic adaptability and estimation error under high-correlation environments. Overall, the findings suggest that regime-based strategies are best interpreted as risk-management tools rather than universal return-enhancing solutions. • GARCH outperforms implied measures regarding regime stability in emerging markets. • Three-regime models capture transitional phases overlooked by binary frameworks. • U.S. regime probabilities act as an exogenous Early Warning System for Brazil. • Dynamic allocation enhances risk-adjusted performance and preserves capital versus static strategies. • High local risk-free rates significantly compress risk premiums in emerging markets.
Bitencourt et al. (Fri,) studied this question.