Satelite production forecasts : vauled with simulated futures and options trading

Abstract

Both the USDA and private firms are allocating substantial capital towards providing accurate and timely crop production forecasts. Production forecasts based on satellite imagery have been suggested as a means of making forecasts earlier, more frequent, and cheaper. This thesis attempts to determine if satellite data increases information with respect to crop condition and final production. If so, does the additional information have value and can it be used to make profitable trades in the futures market? These questions are answered using NDVI data for Iowa and Illinois. Jackknifed out-of-sample crop production estimates are calculated for both corn and soybeans for the individual states. A variety of models were used, each including different bi-weekly periods. USDA crop condition scores are also tested in some of the models. A model based on the current stocks-to-use ratio for each commodity is used to predict the market's expected production level. When the satellite forecast differed from the market's expectation a trade was made in the futures markets. Both futures and option strategies were tested. Results suggest that satellite based production forecasts may result in profitable soybean trades, particularly when downside risk can be reduced by trading options. Further work should focus on refining the satellite images used in the model and exploring more complex option trading strategies.