Forecasting Prices of the Base Metals in the Metals Market

This paper focuses on the construction of two econometric models which will produce reliable out-of-sample forecasts of returns on Zinc and Copper using the prices from London Metal Exchange and various predictors on the sample from January 1995 to March 2017. In order to avoid the problem of non-stationarity first-difference or return of each variable is used. It is shown that the best forecasts are produced by the implementation of Dynamic Model Averaging proposed by Raftery et al. (2010), while relatively simpler models such as GARCH, TGARCH, EGARCH, and ARMA with exogenous variables have considerably higher MSE. However, even the best forecasting DMA has relatively high UII Theil’s coefficient which measures the ratio of MSE of calculated forecasts to MSE of forecasts obtained assuming that the return will not change. Such result is in line with Efficient Market Hypothesis which states that technical analysis cannot provide reliable forecasts. Additionally, the paper shows the change in the determinants of the change in price for each metal with fundamental characteristics such as excess demand and stocks gradually losing importance over time, while US Industrial Production and return on a portfolio of metals-related companies playing bigger role nowadays than in the 2000s.