Narrative Factors: A Dynamic Factor Framework for Risk Management and Alpha Generation

This paper introduces narrative factors, a forward-looking factor framework that maps evolving market narratives into measurable drivers of asset returns. Unlike traditional factor models, which rely on historical return covariation and are inherently backward-looking, narrative factors explicitly model the narratives that shape investor expectations in real time. These narratives include both discrete macroeconomic and event-driven catalysts and continuous shifts in market sentiment reflected in research and investor discourse. By quantifying the evolution, timing, and cross-asset propagation of narratives, the framework makes four contributions. First, it provides a new approach to factor identification that preserves the canonical linear factor model while tying factors to forward-looking catalysts and the evolving expectations surrounding their outcomes, enabling ex ante measurement of narrative risk at the asset level. Second, it formalizes factors as time-specific belief states that evolve as narratives develop and converge when underlying catalysts resolve, offering a more natural representation of how common risks emerge and dissipate in practice. Third, it presents a systematic and scalable method for constructing narrative factors endogenously from unstructured text, without prespecifying themes or relying on static classifications. Finally, by jointly inferring factor states and loadings from narrative evolution and realized returns, the framework supports interpretable risk analysis and generates a rich set of forward-looking signals that can be used for attribution, portfolio construction, and alpha generation.