Caracterização e determinação da eficiência em sistema agrícola de cultivo de soja por meio de análise envoltória de dados e métodos multivariados

Abstract

The development of analysis and management techniques is important. Even if the aim of this work is centered on an energy analysis of a production system, the techniques explored here can be applied in other circumstances. The general objective of the work was to characterize and compare the efficiency of a group of farmers in the region of Boa Esperança, state of Paraná, in the soybean production system using data envelopment analysis and multivariate methods from the perspective of the energy used in the system. The data referring to the amounts of energy used and produced in the systems were converted into energy equivalents per unit of area to perform the energy balance. Then, a bibliographic review on data envelopment analysis (DEA) was carried out. Subsequently, to obtain efficiency values, DEA was used in the set of data obtained from the 27 producers in the 2014/2015, 2015/2016 and 2016/2017 seasons. Multivariate analysis of variance, discriminant logistic regression and clustering analysis in the plane of the principal components were used to verify differences between efficiency classes and clustering pattern considering the different variables. In multivariate analyzes the variables were: 1.productivity, 2.pesticides (fungicides, herbicides, insecticides, adjuvants and spreaders oils), 3.nutrients (N, K2O, P2O5 and micronutrients), 4.seeds and 5.operational (labor, machinery and fuel). The energy balance in production was positive considering the average of 27 producers in three seasons. The total energy used in the production of soybeans was 40,888.98 MJ ha-1. A balance of 53,634.69 MJ ha-1 was found, debiting equivalent energy expenditures from energy equivalent productivity. The ratio between the totals of output and inputs was 2.31. It was found in a mathematical approach that it is possible to obtain efficiency values by building relative efficiency values between elements of a group of decision making units (DMU’s). Using DEA with input orientation in the 27 soybean producers in the 2014/2015, 2015/2016 and 2016/2017 harvests, the maximum efficiency value was 100% and the lowest value was 71.37%. Two classes were defined: “efficient” and “inefficient”. The mean vectors associated with the “efficient” and “inefficient” classes show a significant difference by the Wilks Lambda test at the level of 5% according to the multivariate analysis of variance. Simultaneous intervals with at least 95% confidence estimated by the Bonferroni method for the difference between the mean vectors indicate that the means of the “operational” and “productivity” variables are different. The estimated logistic regression was able to distinguish between the “efficient” and “inefficient” classes in more than 75% of the total cases evaluated. In the clustering analysis by principal components, three clusters were identified considering the average of clusters. The balance between energy cost and productivity per unit area characterizes cluster 1. Cluster 2 is characterized on average by productivity values in low energy equivalents and high expenditure of energy equivalents with pesticides and operational. Cluster 3 has efficient and inefficient DMU’s. The DMU's in this group have high (low) productivity associated with high (low) energy expenditures.