Forecasting energy commodity prices: A large global dataset sparse approach

This paper focuses on forecasting quarterly nominal global energy prices of commodities, such as oil, gas and coal, using the Global VAR dataset proposed by Mohaddes Raissi (2018). includes a number potentially informative macroeconomic variables for 33 largest economies, overall accounting more than 80% GDP. To deal with information this large database, we apply dynamic factor models based penalized maximum likelihood approach that allows to shrink parameters zero estimate sparse loadings. The estimated latent factors show considerable sparsity heterogeneity in selected loadings across variables. When model is extended predict commodity up four periods ahead, results indicate larger predictability relative benchmark random walk 1-quarter ahead all commodities 4 quarters prices. Our also provides superior forecasts machine learning techniques, elastic net, LASSO forest, applied same database.