Abstract: Complete replication and data package. This paper proposes methods to include information from the underlying nominal daily series in model-based forecasts of average real series. We apply these methods to forecasts of the real price of crude oil. Models utilizing information from daily prices yield large forecast improvements and, in some cases, almost halve the forecast error compared to current specifications. We demonstrate for the first time that model-based forecasts of the real price of crude oil can outperform the traditional random walk forecast, that is the end-of-month no-change forecast, at short forecast horizons.

Note that Yahoo Finance data in MasterFile_CDataM.csv is omitted from column AE; data may be sourced from link provided in Related Items section.

Replication package for peer-reviewed article published in International Journal of Forecasting. Paper published online March 14, 2025. When citing this dataset, please also cite the associated article. A sample Publication Citation is provided below.