Noisy market, machine learning and fundamental momentum

Abstract

We employ machine to learn the continuous fundamental information and elucidate the fundamental momentum in the noisy Chinese stock market. We extract fundamental implied component from realized returns and sort stocks with the trend of implied parts. The high-dimensional fundamental momentum significantly differentiates from its predecessor, yielding an annualized return of 13.8%. Underreaction in investors helps explain the strategy, as it generates stronger profit during periods of low investor sentiment and in subsamples with high idiosyncratic volatility. The retail investors in China are prone to distort the presentation of momentum. Fundamental momentum is robust in the U.S. samples, different training windows and alternative machine learning algorithms.