Forecasting excess stock returns with crude oil market data

• Oil market information has predictive ability to stock returns.
• A dynamic model selection approach is used.
• The forecasting accuracy can be improved by allowing for parameter change.
• The forecasting gains are significantly in both economic and statistical sense.

In this paper, we forecast excess stock returns of S&P 500 index from January 1997 to December 2012 using both well-known traditional macroeconomic indicators and oil market variables. Based on a dynamic model selection approach, we find that the forecasting accuracy can be improved after adding oil variables to the traditional predictors. The forecasting gains relative to the benchmark of historical average are statistically and economically significant. Moreover, time-varying parameter models generate more accurate forecasts than constant coefficient models.