Forecasting pseudo-periodic seasonal patterns in agricultural prices

摘要

A seasonal model is proposed to forecast agricultural prices with pseudo-periodic seasonal patterns, in which the length of the seasonal period does not remain the same over time. The seasonal effect at a season is defined as a function of the proportion of the seasonal period length elapsed up to that season, and the seasonal pattern is modeled by means of evolving splines to capture any dynamic process of change. Such a model is a useful tool to forecast seasonal behaviors. To illustrate the relevance of this modeling framework, the methodology is applied to weekly prices of tomatoes exported to German markets.