Accounting for rainfall and the length of growing season in technical efficiency analysis

The physical environment of farming systems is rarely considered when conducting farm level efficiency analysis, which is likely to lead to bias of performance measurements based on benchmarking methods such as Data Envelopment Analysis (DEA). We incorporate variations of the physical environment (rainfall and length of growing season) through the specifications of the linear programming in DEA to investigate performance measurement bias. The derived technical efficiency estimates are obtained using a sub-vector DEA which ensures farms are compared in a homogenous environment (i.e. accounting for differences in rainfall levels amongst distinct farm units). We use the Farm Business Survey to analyse a representative sample of 245 cereal farms in the East Anglia region between 2009 and 2010. Efficiency rankings obtained from a standard DEA model and a non-discretionary DEA model that incorporates the variations in the physical environment. We show that incorporating rainfall and the length of the growing season as non-discretionary inputs into the production function had significantly altered the farm efficiency ranking between the two models. Hence, to improve extension services to farmers and to reduce biased estimates of farm technical efficiency, variations in environmental conditions need to be integral to the analysis of efficiency.