Siddhartha Bora, a Ph.D. student in the Department of Agricultural, Environmental, and Development Economics (AEDE), has been awarded the 2021 William E. Krauss Director’s Award for Excellence in Graduate Research for his lead-authored paper entitled "The Rationality of USDA Forecasts under Multivariate Asymmetric Loss". The award is given by the OSU College of Food, Agricultural, and Environmental Sciences to a current or recently graduated Ph.D. student that has produced a peer reviewed publication of the highest quality in the past year. Siddhartha’s paper is published in the American Journal of Agricultural Economics (AJAE), the flagship journal in the field of agricultural economics with an impact factor of 3.028. This paper is a part of his dissertation work supervised by Dr. Ani Katchova and supported by the Farm Income Enhancement Program.
Siddhartha’s study builds upon previous findings of underprediction bias in the USDA forecasts by several studies. The bias toward underprediction has been especially prominent in the early farm income forecasts. These prior results seem to suggest that USDA may not want to paint a rosy picture of the farm economy in the beginning of the year. However, as the year progresses, USDA forecasters revise their forecasts, and these revised forecasts tend to be less biased. This evidence of an under-prediction bias in the USDA forecasts raises the question whether the USDA forecasters are “rational” in producing such forecasts.
Forecasters may produce biased forecasts but may yet be “rational” if they are minimizing their loss due to forecast errors. In his paper, Siddhartha evaluates the “rationality” of two major series of USDA forecasts, the farm income forecasts released by the Economic Research Service and the World Agricultural Supply and Demand Estimates (WASDE) price and production forecasts produced by the World Agricultural Outlook Board. These forecasts have important implications for farm policy, as they are frequently mentioned in policy debates in the media and in Congress. These forecasts are also very important for commodity markets, and previous studies have shown that WASDE announcements significantly impact commodity futures prices. Both the farm income and WASDE forecasts are heavily cited statistics produced by the USDA and are eagerly awaited by farmers and stakeholders in the agricultural sector.
Using a novel multivariate asymmetric framework, Siddhartha shows that USDA forecasts are rational under the assumption that USDA places different weights on under- and over-prediction errors. For example, in the case of net cash income forecasts, the USDA forecasters tend to under-predict the actual net cash income since they have a higher cost associated with over-prediction errors. This approach is different from traditional studies, which typically do not differentiate between under- and over-prediction errors by minimizing a mean square (MSE) loss. This novel framework provides a new direction to similar lines of research on forecast bias.
Another contribution of his study is that the joint evaluation of a series of forecasts of several variables. This makes sense since the USDA releases several variables as part of the same report. Evaluating each variable separately might ignore any interactions between them. Finally, while the results are mostly consistent with previous findings, they yield an alternative interpretation of the results of prior research. These findings shed light on “why” the USDA forecasters may produce biased forecasts.
This research on the rationality of the USDA forecasts is very timely given the recent economic downturn in the agricultural sector. Due to the recent pandemic and resulting uncertainty about farm income, and the farm economy in general, USDA forecasts have received additional scrutiny from a variety of forecast users. The study addresses questions that are relevant to a wide range of players in the agricultural sector who use the USDA forecasts, including farmers, lenders, agricultural business leaders, and agricultural policymakers. Another implication of the findings is that they may help inform future revisions of USDA forecast models and procedures. Siddhartha continues to pursue his interests in farm economy, and he is currently investigating the accuracy of long-run agricultural baseline projections for the US farm sector in a separate study.
References:
Bora, S.S., A.L. Katchova, and T.H. Kuethe. “The Rationality of USDA Forecasts under Multivariate Asymmetric Loss.” American Journal of Agricultural Economics (2020), https://doi.org/10.1111/ajae.12142