We decompose the idiosyncratic volatility of stock returns into “good” and “bad” volatility components, which are associated with positive and negative returns, respectively. Using firm characteristics, we estimate a cross-sectional model for the expected idiosyncratic good minus bad volatility (EIGMB). The EIGMB outperforms expected idiosyncratic skewness (EISKEW) and standard time-series models in capturing conditional idiosyncratic return asymmetry. EIGMB is negatively and significantly associated with future stock returns, even after controlling for EIKSEW and exposure to systematic-skewness-related factors. Separating the role each specific characteristic plays in driving the predictive power of EIGMB for returns, we find that return on equity and momentum are two important elements of variation in EIGMB.

Binsbergen, Han, and Lopez-Lira (2023) predict analysts’ forecast errors using a random forest model. A strategy that trades against this model’s predictions earns a monthly alpha of 1.54% ($ t $-value = 5.84). This estimate represents a large improvement over studies using classical statistical methods. We attribute the difference to a look-ahead bias. Removing the bias erases the alpha. Linear models yield as accurate forecasts and superior trading profits. Neither alternative machine learning models nor combinations thereof resurrect the predictability. We discuss the state of research into the term structure of analysts’ forecasts and its causal relationship with returns.