Analyzing the impact of recent developments in agricultural commodity markets, and using machine learning to improve forecasting techniques

USDA reports are highly informative and valued by market participants. Markets quickly absorb their information, and exhibit volatility spikes following report announcements that dissipate quickly (Adjemian and Irwin ­ 2018). A government shutdown in January 2019 prevented the USDA from publishing information about the situation and outlook for major U.S. agricultural commodities. I show that, as a result, Chicago Mercantile Exchange Board of Trade markets for corn and soybeans experienced heightened market uncertainty, elevating the cost of managing risk using options. Having established the importance of government information, I decompose ending stocks forecast errors into errors of the other supply and demand components using a Machine Learning (ML) algorithm. The results show that export and production misses are the major contributors to ending stocks projection errors. Lastly, using past yield shocks as instruments to endogenous futures prices, I estimate a supply elasticity for corn at 0.26 percent. Recently, due to the COVID-19 pandemic, demand for ethanol plummeted – resulting in an inward shift in the demand curve for corn. Consequently, my analysis shows that corn producers suffered losses worth $5.4 billion, 27% lower compared to the actual payments USDA made under Coronavirus Food Assistance Program (CFAP). Moreover, all my estimates suggest that the USDA was able to fully compensate producers for any losses they faced due to demand reduction.