Analysis of efficiency gains in commodity price forecasting using ARMA models with varying levels of data aggregation

This study evaluates the efficiency gains in forecasting three commodity prices (live cattle, coffee and cotton) using different ARMA models with varying levels of temporal aggregations (weekly, monthly, quarterly and annually). More specifically, it evaluates whether the disaggregated models can produce more accurate aggregated price predictions than the corresponding aggregated models. The commodity prices were adjusted using the Consumer Price Index (CPI). Likewise, they had a non-stationary behavior and heteroskedasticity issues; therefore they were subject to detrending and transformations using GLS. Under the three different scenarios, disaggregation levels effectively provided an efficiency gain in forecasting, and the best models were the weekly models. The same behavior was consistent across all possible levels of aggregations, i.e. monthly models had a better performance than quarterly and annual models in forecasting quarterly and annual prices