Uncovering Unexploited Fundamental Signals for Stock Selection: A Systematic Data Mining Approach with Bootstrapping

Abstract

This study implements a systematic data-mining approach to uncover unexploited fundamental signals for predicting the cross-section of stock returns in the Japanese stock market. Specifically, we construct a universe of about 360,000 signal portfolios by systematically combining two fundamental signals from all available financial statement items. Using a cross-sectional bootstrap methodology, we uncover a number of fundamental-based return predicting signals beyond the popular Fama-French factors. Our evidence suggests that the return predictability of these signals cannot be attributed solely to random chances (false discoveries). Some of the uncovered fundamental signals involve various measures of corporate liabilities, profits, and shareholder payouts.