Forecasting Sector Stock Market Returns

Abstract

We seek to forecast sector stock returns using established predictor variables. Existing empirical evidence focuses on market level data and thus sector data provides fertile ground for research. In addition to in-sample predictive regressions, we consider recursive and rolling forecasts and whether such forecasts can be used successfully in a sector rotation portfolio. The results for ten sectors and eleven predictor variables highlight that two variables, the default return and stock return variance, have significant predictive power across the stock market series. Forecast results are also supportive of these series (especially the default return), which can outperform benchmark and alternative forecast models across a range of metrics. A sector rotation strategy based on these forecasts produces positive abnormal returns and a Sharpe ratio higher than the baseline model. An examining of the sectors at each rotation reveals that a small number of dominate in the constructed portfolios.