A Filtering Strategy for Improving Charateristics-Based Portfolios

Abstract

In this paper, we propose new indexes to measure the predictive power for future returns possessed by firm characteristics and find that the predictive power significantly differs across the cross-section of assets. We also propose a filtering strategy to improve conventional characteristics-based portfolio profits. The new strategy filters out assets with low predictive power. We apply the new strategy to equity data and find that it significantly outperforms the conventional strategy for several well-known firm characteristics. We also find that characteristics-based portfolio profits are not prevalent but rather driven by only a small subset of stocks.