Forecasting crude oil prices: do technical indicators need economic constraints?

This study aims to improve the forecasting performance of technical indicators for crude oil prices by imposing economic constraints on the sign of the shrinkage estimators. The out-of-sample results indicate that our constrained methods deliver significantly stronger forecasts than their standard forms and prevalent predictive models. Moreover, the advantages of the constrained methods are stronger during recessions and weaker during expansions. The superior forecasting performance of the constrained methods is not affected by the consideration of long-horizon forecasts and is robust to a large body of alternative specifications. In addition to the statistical tests, we provide evidence that investors who use the new predictive framework can realize sizable economic gains through asset allocations and market timing.