Cotton Price Forecasting and Structural Change

Agricultural prices have long been forecast with reduced-form models including ending stocks as an independent variable. In recent years, cotton prices have been persistently low compared with the other agricultural products that compete with cotton for land and other inputs. Furthermore, the cotton price forecasting models used by a number of entities have chronically realized positive errors—persistently forecasting prices too high. This paper reviews some general principles behind short-term agricultural price forecasting, discusses some of the issues specific to specifying cotton price forecasting models, and compares the forecasting performance of an number of alternative specifications. The discussion and results are intended to lay the basis for developing new models that account for structural changes in world cotton markets.