Stock Return Predictability Based on a Latent Factor Model Perspective: Evidence from China A‐Shares*

Abstract This study investigates the predictability of cross‐sectional stock returns in Chinese A‐share market. We employ an instrumented principal component analysis (IPCA) approach on a wide range of firm‐level characteristics to construct a dynamic 4‐IPCA index as a novel predictor. We find that this index has strong predictive power, with a long‐short strategy yielding a monthly risk‐adjusted alpha of 190 basis points. Mechanistically, this predictability is more pronounced for firms receiving less investor attention, supporting a mispricing‐based explanation, while lottery preferences and arbitrage limitations hold no explanatory power. Furthermore, the 4‐IPCA effect is significantly associated with aggregate market conditions, including investor sentiment, market development, trading frictions, and accounting quality.