Research

Abstract: This paper develops a framework that connects macroeconomic risks to stock factor returns by extracting latent return factors with the largest overlap to the space of macroeconomic risks. I test this framework on the dataset of hundreds of macroeconomic series and factor portfolio returns. The findings of the test are twofold. First, the extracted return factors have higher prices of risk than other benchmark return factors. Second. the extracted macroeconomic factors have better predictive powers, both in-sample and out-of-sample, than individual macroeconomic series. These results provide guidance for identifying the macroeconomic risks related to empirical asset pricing research.

Abstract: Existing PCA methods identify common risk factors while ignoring information about expected returns. We propose a general and flexible moment-fusing approach that incorporates risk prices directly into the identification of risk factors. Specifically, we construct a return transformation such that the volatilities of the transformed returns are proportional to the Sharpe ratios of the original assets. We then perform factor analysis on the transformed returns, with or without additional pricing restrictions. Empirically, we find that the moment-fusing factors significantly outperform leading benchmark models in pricing both FX and equity portfolios out-of-sample.

Abstract: We model the dynamics of expected returns for a large set of long-short portfolios based on characteristics from the return anomaly literature. Our models permit both cyclical forms of expected return variation and permanent decay effects. We document statistically and economically significant cyclical variation in anomaly portfolio expected returns. From an ex-post perspective, the majority of historical variation in expected anomaly portfolio returns is attributable to the cyclical component, rather than permanent decay effects. The most successful predictors appear to be the value spread, measures of anomaly portfolio momentum, and equity market sentiment. We emphasize the value of pooling information across anomalies via panel predictive regression models. Such models both clarify the evidence for predictability and generate out-of-sample forecast improvements relative to anomaly-specific forecasting approaches.

Abstract: This paper examines how heterogeneous learning speeds about inflation among investors contribute to cross-sectional variation in stock returns. I develop an asset pricing model in which investors update inflation expectations through Bayesian learning, which leads to persistent belief heterogeneity that impacts firm valuation. The model, supported by an empirical illustration, shows that return differentials widen when learning disparities increase. It further predicts that when slow learners dominate, pricing bias becomes more pronounced even as forecast errors remain limited. These results underscore the role of inflation expectation formation in shaping return dynamics and cross-sectional mispricing.