Mispricing and Correction in Short-Term Returns

Abstract

Our study addresses a limitation of the traditional short-term reversal strategy, which presumes that all stocks have the same expected return and identifies mispricing solely based on realized returns. We estimate the conditional expected return of a stock using firm characteristics and machine learning and measure the mispricing of a stock as Short-Term Excess Return (STER), the difference between its realized return and conditional expected return. Our results show that machine learning methods produce more accurate conditional expected returns and that a STER-based reversal strategy generates significantly higher returns than the traditional short-term reversal strategy. Our findings are consistent with the investor overreaction-based explanation.

Keywords: Mispricing, Short-Term Reversal, machine learning, Asset Pricing