Timing the Market with Valuation and Trend

Abstract

A market timing strategy, based on a non-linear forecasting model, is shown to deliver economically significant returns over a buy and hold strategy. The model is developed on the S&P 500 index using macro economic variables. The distribution of returns has a conditional mean that can be determined ex-ante. Using the regime given by the classification model, a market timing strategy is implemented on out-of-sample data from July 1990-August 2004. The results indicate that investors can and should condition their investment strategy on the basis of their information set. However, in order to adopt such a strategy, it may be necessary to use a nonlinear forecasting technology.