Explaining the Idiosyncratic Volatility Puzzle With a Bayesian-Updating Model

Abstract

We document that the idiosyncratic volatility (IVOL) puzzle exists only among firms that underperform their benchmark or release negative earnings surprises. We explain these findings using a Bayesian updating model with asymmetric signal precision in which agents observe noisy signals about future cash flows. In this setting, negative news are associated with relatively lower signal precision, negative momentum, and low subsequent returns. After controlling for relative performance (our proxy of news) and signal precision, the IVOL puzzle disappears. This performance- and signal-precision-based explanation alone can account for up to 83% of the IVOL puzzle, which is more than other existing theories combined.