This study evaluates all USDA cotton supply and demand estimates for the United States and China (including unpublished price forecasts) from 1985/86 through 2009/10 for accuracy and efficiency. Results reveal that at every stage of the forecasting cycle forecast smoothing was the most widespread and persistent type of inefficiency observed in most U.S. variables. Correlation with past errors indicated the tendency to repeat past errors in most cases. Tendency to overestimate growth was also found. Bias was uncommon and limited to several cases of overestimation of China’s exports and U.S. price and underestimation of China’s domestic use. While forecasts of China’s imports and endings stocks improved, U.S. price and ending stock forecast errors became larger toward the end of the study period.