MODELING ANNUAL AND QUARTERLY U.S. FARM TRACTOR SALES
thesisposted on 23.04.2020, 18:53 by Kylie M O'ConnorKylie M O'Connor
Farm machinery is a vital input for production agriculture and, as a result, is a significant part of the agricultural economy. Despite its great importance, there has been relatively little academic analysis on the driving forces behind farm machinery sales over the past several decades. The studies that do evaluate farm machinery sales all do so regarding annual sales despite shorter-term sales data being available. These previous studies primarily use traditional macroeconomic variables, tailored to the agricultural industry, to explain farm machinery sales. Recently, with the creation of the Ag Economy Barometer Survey in October 2015, farmer sentiment data is being collected. Studies using consumer sentiment data to evaluate consumer demand have found sentiment data useful when including it in demand models, especially for consumer durable goods. This study evaluates farm machinery sales, specifically two-wheel-drive tractors with 100 horsepower or higher, using both traditional macroeconomic variables and farmer sentiment data. The evaluation begins by looking at annual tractor sales from 1978 to 2019 using machinery prices, prices received for outputs, prices paid for inputs, lagged net farm income, interest rates for loans specifically for farm machinery, farm assets, and the number of acres harvested. The annual models are used to derive elasticities with respect to farm tractor sales, and the quantity demanded is most responsive to changes in machinery prices, the number of acres harvested, prices received for crops and livestock, and the level of farm assets. Out-of-sample estimations aids in evaluating the forecasting power of the models with the best statistical fit. The model with the best out-ofsample performance forecasts 2020 sales of farm tractors with 100 HP and above using various assumptions for agricultural economic conditions in 2020. The model estimates a record low in tractor sales dating back to 1978. The annual models are then re-estimated using quarterly data spanning from 2009 to 2019. The quarterly models have less statistical fit than their annual counterparts. This reduced model performance is likely due to the seasonal nature of farm tractor sales and that some of the explanatory variables are only updated on an annual basis, limiting their ability to capture the seasonal variations. Finally, the quarterly models are estimated again to include farmer sentiment data. At the time of the study, only 17 quarterly observations of farmer sentiment data had been collected, significantly limiting the evaluation. The limited number of observations results in an inconclusive outcome regarding the explanatory power of farmer sentiment data.