A statistical model to forecast season averages of international cotton prices has been identified by the ICAC Secretariat. The 2007 model explains 88% of the annual variation since 1975 in the Cotlook A Index, an indicator of international prices.
The 2007 model succeeds an earlier model first developed in 1988 and revised several times. The 2007 model will assist the Secretariat in developing improved forecasts of cotton prices.
A report detailing the econometric work and data used in developing the ICAC 2007 Price Model is available at ICAC website and is being published in the July/August issue of COTTON: Review of the World Situation.
Net trade in cotton by China (Mainland) was used as an explanatory variable in the 1988 ICAC model, and as imports by China (Mainland) rose in recent seasons, the prices indicated by the model were substantially above realistic levels in 2005/06 and 2006/07.
The 2007 model relates changes in the Cotlook A Index to changes in the stocks-to-mill use ratio in the “World-less-China” and to past changes in the stocks-to-mill use ratio in the “World-less-China” and “China”. Splitting the world into two regions (“China” and “World-less-China”) provides better explanatory power than considering the world as one region.
According to the 2007 model, a 5% increase in the stocks-to-mill use ratio in the World-less-China results in a decrease of 4.9% in the season-average Cotlook A Index during the same season.
If the stocks-to-mill use ratio in the World-less-China continues to increase by another 5% in the following season, then the combined effect of the two-season increase in the stocks-to-mill use ratio in the World-less-China is expected to reduce the season-average Cotlook A Index by 6.6%.
The stocks-to-mill use ratio in China has a lagged effect on the Cotlook A Index. When the China stocks-to-use ratio increases by 5%, the average Cotlook A Index in the next season decreases by 0.36%.
To evaluate how well the model would have predicted prices in the past, one-season-ahead price forecasts were simulated with information at the end of each season for the seasons 1990/91 to 2006/07.
With perfect foresight of the stocks-to-mill use ratios in the World-less-China and China, the model would have correctly predicted price increases and decreases in all cases. The average absolute difference between the forecast with perfect foresight and the observed Cotlook A Index amounts to 4.8 cents per pound.
On the other hand, if historical forecasts of the stocks-to-mill use ratio are used instead of the actual ratios, price increases and decreases would have been correctly forecast 11 out of the 17 seasons, and the average absolute difference between the historical forecast and the actual Cotlook A Index amounts to 9.5 cents per pound.
Beginning in 2007/08, forecasts of the Cotlook A Index will be revised monthly based on a weighted average of the observed values of the Cotlook A Index during the current season and the latest annual forecast for the remainder of the season.
