Predicting Stock Market Returns with Time-VaryingModels and Parameters

Abstract

This article investigates the controversial issue of stock market return predictability by using commonly used regression methodology and a parsimonious set of business cycle variables that were identified by the 1980s. Unlike earlier studies, the authors treat both models and estimation windows as variables. Their findings provide evidence of out-of-sample stock market return predictability related to business cycles. The results have practical utility for investment practitioners because their optimal models, based on a methodology, variables, and data that were available for real-time forecasts, provided significant explanatory power for future monthly stock market returns; their forecasts delivered substantial utility gains compared with average historical returns; and investment strategies based on the forecasts improved risk–return trade-offs compared with buy-and-hold stock market investments.