We find remarkably low correlation among major investor disagreement proxies, making it challenging to infer how disagreement impacts prices. We develop a unified model that incorporates these measures into a novel nonlinear composite disagreement measure, which is more predictive of returns than existing measures, and that explains the stronger disagreement-expected return effect for smaller stocks. Decile spread portfolios sorted on our composite measure generate value-weighted alphas of 14.2% per year and 20.5% among smaller stocks. Our model also provides a disagreement-based rationale for the idiosyncratic volatility puzzle, which is absorbed by our composite disagreement in empirical tests.

A factor using an earnings measure treating intangible and physical investments symmetrically represents “quality.” It has smaller left tail risk and co-tail risk with the market than does RMW of Fama and French (2015) and has lower down-market than up-market exposure. Our factor has significant alpha relative to many extant multi-factor asset-pricing models, including the Fama-French model (α = 2.9%). Its performance is due to superior asset selection (market timing) on the long (short) side. When the profitability factor in the Fama-French model is replaced with our factor the resulting model performs better in explaining both the cross section of stock returns and several extant anomalies.

Sell-side analysts' forecasts of stock returns are biased and the consensus forecast is a poor cross-sectional predictor. In sharp contrast, the implicit ranking of stocks by each analyst is highly informative of subsequent returns. Long-short portfolios sorted on these rankings result in large and highly significant excess returns that cannot be explained by previous anomaly characteristics. The strong performance is most easily understood by noting the similarity between rankings and within-analyst demeaned forecasts. The latter are equivalent to removing each analyst's fixed effect and thus controlling for unobservable analyst-specific biases. We document analogous results using analysts' recommendations and earnings forecasts.