Prediction of Stock Returns Based on Cross-Sectional Multivariable Model

Abstract

A new prediction method of stock returns was constructed from a cross-sectional multivariable model where explanatory variables are current financial indexes and an explained variable is a future stock return. To achieve precise prediction, explanatory variables were appropriately selected over time based on various test statistics and optimization of a performance index of expected portfolio return. A long-short portfolio, in which stocks with high predicted return were bought and stocks with low predicted return were sold short, was constructed to evaluate the proposed method. The simulation test showed that the proposed prediction method was effective to achieve high portfolio performance.