Regional Commodities Price Volatility Assessment Using Self-driven Recurrent Networks

PABLO NEGRI; RAMOS, MARIA PRISCILA; BREITKOPF, MARTIN

Porto

Congreso; 25th Iberoamerican Congress on Pattern Recognition; 2021

IAPR

The high volatility of the agricultural and energy commodity prices in the international market is a concern due to their transmission to regional prices, increasing instability in domestic markets. This paper evaluates the performance of recurrent networks (RNN and LSTM) to predict regional prices reactions under international shock simulations. Experiments are run to soybean and corn regional prices in Argentine by considering exogenous changes of the international oil price - both agricultural commodities are inputs for biofuels' production - and also of their international prices. Results are in line with the econometric literature and consistent with the dynamic of regional prices in Argentina'smarkets. Thus, the RNNs could be a useful tool for timely economic policy decisions that cushion external price shocks in domestic markets.