Are Disagreements Agreeable? Evidence from Information Aggregation
60 Pages Posted: 4 Dec 2017 Last revised: 20 Jun 2019
Date Written: June 1, 2019
Disagreement measures are known to predict cross-sectional stock returns but fail to predict market returns. This paper proposes a partial least squares (PLS) disagreement index by aggregating information across individual measures and shows that this index significantly predicts market returns both in- and out-of-sample. Alternative machine learning methods not only confirm the predictability of disagreement on market returns but also identify the most important disagreement measures. Consistent with Atmaz and Basak (2018), the predictability of our disagreement index is stronger in high sentiment periods, operates via a cash flow channel, and extends to market volatility and trading volume.
Keywords: Disagreement, Return predictability, PLS, PCA, LASSO, Machine learning
JEL Classification: G12, G14
Suggested Citation: Suggested Citation