Narrative Asset Pricing: Interpretable Systematic Risk Factors from News Text.

This article presents a novel empirical asset pricing model that estimates Intertemporal Capital Asset Pricing Model (ICAPM) state variables and risk factors using business news narratives from The Wall Street Journal. The methodology integrates latent Dirichlet allocation (LDA) for topic modeling, instrumented principal component analysis (IPCA) with a group lasso penalty (Sparse IPCA) for factor extraction and variable selection, enabling the identification of a parsimonious set of narrative-based risk factors. These narrative factors outperform traditional characteristic-based factor models in explaining asset returns, achieving higher out-of-sample Sharpe ratios and smaller pricing errors, while also exhibiting predictive power for future macroeconomic variables consistent with ICAPM theory. The model further allows interpretation of risk factors by linking them to specific news topics and terms, with narratives such as "Recession" having the largest negative impact on the pricing kernel. This approach demonstrates the potential of using textual news data to capture fundamental economic risks relevant for asset pricing.