This study provides a comprehensive examination of accuracy and efficiency of all USDA cotton supply and demand estimates for the U.S. (including unpublished price forecasts), China and rest of the world (ROW) over1985/86 through 2009/10. Our findings show that USDA overestimated China’s exports and underestimated China’s domestic use and ROW imports. Based on correlation of forecast errors with levels, estimates of U.S. domestic use, ending stocks and China’s exports were too extreme while forecasts of China’s ending stocks and ROW production and exports were too conservative. Correlations with past errors suggest that USDA tends to repeat errors in ROW production forecasts and overcorrect errors in ROW exports forecasts. Significant positive correlation between subsequent revisions indicating forecast “smoothing” was detected in the U.S. production, domestic use, exports and ending stocks forecasts, China’s imports, domestic use and exports forecasts and the ROW production and domestic use forecasts. While China’s ending stocks and production forecasts significantly improved over time, (unpublished) U.S. price forecasts became worse. Based on correlations of errors we conclude that better forecasts of U.S. ending stocks and domestic use forecasts, China’s imports and ROW ending stocks and exports forecasts are essential for improving U.S. cotton price forecasts.