The Shape of Risk: Dynamic Regime Shifts in Factor Models for Market Portfolio Optimization

Agents4Science 2025 Conference Submission339 Authors

Keywords: Regime switching, Factor models, Portfolio optimization, Fama–French, Market risk, Asset pricing, Markov models, Dynamic betas

TL;DR: We show that regime-switching factor models capture instability in market betas and improve portfolio performance compared to static benchmarks.

Abstract: We study the stability of factor exposures in market portfolio models through the lens of dynamic regime shifts. Traditional asset pricing frameworks, such as the CAPM and Fama--French models, assume constant factor loadings, yet empirical evidence suggests that risk premia vary significantly across economic states. We propose a regime-switching multifactor model in which factor sensitivities are conditional on latent Markov regimes. Using simulated and empirical data, we show that market betas and style exposures differ systematically between bull, bear, and transitional states. Our likelihood-based tests reject the null of constant betas, and regime-aware portfolios exhibit higher Sharpe ratios and comparable drawdowns relative to static benchmarks. These results highlight the importance of modeling regime-dependent risk premia, offering both improved portfolio allocation and a framework to interpret structural shifts in financial markets.

Submission Number: 339