Beschreibung:
This paper aims to investigate the predictability of crude oil futures prices using a data-heavy approach with traditional macroeconomic and financial variables as well as new predictors. To do so, we adopt diverse combination approaches of bivariate regressions. A particular emphasis of the paper is on the incentive of addressing Stambaugh's bias for forecasting crude oil. We find that this is a minor issue compared to the choice of the forecasting model. We also show that exchange rates and technical indicators are somewhat favored predictors of crude oil. Empirical findings reveal that blending PCA with soft-thresholding selection technique can yield particularly good results in small samples