A study about the performance of time series models for the analysis and forecasting of agricultural prices

Abstract

Agribusiness is simply the biggest deal of the Brazilian and world economies. The price is an important variable for this sector since the economic and market conditions have been varied over time. Efficient modeling methods have been needed in order to better describing trends and characteristics of that variable. The objective was to study the performance of time series techniques of fixed and open models for the analysis and forecasting of agricultural prices, in order to demonstrating the feasibility of those techniques in generating results for making economic decisions. Historical series of monthly average price of the crops peanuts, sugar cane, banana and orange received by Brazilian producers were considered in the analysis, from August 1994 to December 2009. The study showed that although exponential smoothing models to be not complex, they performed better in the adjustment of most price series.