Dr. Carl Anderson passed away August 30, and we bade him farewell a few days before this column was written. He was a fine gentleman, very generous with his time and talents. One of his talents was forecasting markets. Others have already summarized Carl’s achievements better than I ever could. What I would like to do in this column is highlight his approach to forecasting cotton prices. This column is partly in memorium, but also serves as a teaching example. Carl liked good teaching examples.
Carl was a traditional economist trained to think about the world in terms of supply and demand. His emphasis was always on fundamental elements that resulted in a balance of supply and demand (or an excess of one over the other). He was experienced enough to know about the influences and distortions that government policies and, more recently, the fund sector, could have on commodity markets.
For the latest on southwest agriculture, please check out Southwest Farm Press Daily and receive the latest news right to your inbox.
Carl was also of the old school when it came to his style of analysis. He would gather data – LOTS of data. He could afford to do that because his office associate of 35 years is an exceptionally talented database developer and manager. After collecting all the publicly available data there was to be had, and having it organized into tables and charts, he would ponder and mull over it for a while. He would use his numerous contacts in the industry to confirm or update trends he saw in the data. And then he would develop a conservative ad hoc forecast of the possible outcomes for supply, demand, ending stocks, and ultimately prices. By ad hoc, I mean he used his own thought processes, sometimes mixed with his gut feelings. By conservative I mean that he applied the most basic form of risk modeling: using best/worst/most likely (in his opinion) outcomes for the fundamental variables in question.
The more modern way of forecasting is to use more sophisticated statistical and computer tools to project prices and model risk. Carl would certainly consider the predictions made by statistical modelers in developing his ad hoc forecast. He recognized that statistical approaches have advantages and disadvantages. He collaborated in a cotton price modeling project for a number of years at Texas A&M. But ultimately his approach was old school.
In the end, Carl was a successful analyst because he was pretty accurate in his longer term forecasts of the cotton market. One example sticks out in my memory. In early January, 2004, many of us were in San Antonio attending the Beltwide Cotton conference. The cotton market had rallied up to 80 cents during the previous fall, and was then (in January) trading in the mid-60s. There were differing opinions about whether the market would revisit 80 cents or get weaker. Carl had the latter viewpoint, given his fundamentally-oriented notion that higher prices in the winter would reduce consumption, increase supply, and build ending stocks. I recall reading his newsletters and giving some Extension presentations that emphasized his caution about the possibility of lower prices. As it turned out, Carl’s forecast was dramatically correct. Cotton prices fell into the 40-cent range, trumping any previous technical indications in the first quarter of that year.
Carl’s view of the 2014 crop was similarly focused on the possible outcomes of world cotton consumption, Chinese import demand, U.S. (and especially Texas) production, and Indian production. We will all have a slightly clearer notion of those variables after the September report by USDA, but it will likely take several more months of data to confirm the picture.