Quebra de safra de commodities agrícolas utilizando a abordagem de modelos Probit dinâmicos

Abstract

This study aimed to estimate the probability of crop failure of the main agricultural crops in Brazil considering a wide set of covariates and using different classes of dynamic Probit models. A crop break schedule was implemented, considering three criteria: chronology of events by crop and federative unit, variation in production and number of insurance triggers. Considering the Probit models: static, dynamic,auto-regressive and dynamic auto-regressive, the selection of variables using the Stepwise method was implemented as a first step, which made it possible to form a set of covariables to estimate the probability of crop failure occurring of soy, corn, cotton and wheat crops. The results showed that the Probit models, intheir different variants, presented percentages of correct answers above 70 % in all cultures, however they also presented false-positive rates above 10 %. Wheat had the highest average probability of crop failure, with soybean having the lowest probability of crop failure. Commercial activity, the number of insurance policies and climatic conditions were the most relevant variables in the estimated models. Wheat cultivation is more affected by economic conditions than by climatic conditions, with a migratory effect of areas planted between soy and wheat being identified in this model, while cotton was more affected by a greater set of climatic variables than economic variables. Second crop corn (safrinha), being the largest corn crop, is most affected by internal economic conditions while first crop corn (verão) is most affected by external economic conditions, although it was found in this study that crop failure uncertainty, measured by error -standard of the estimated probability, is higher for second crop corn. Soy despite having a good fit within the sample and with the largest number of selected variables, presented the highest estimated standard error for crop failure probability, this being explained by the presence of both climatic variables and variables of internal economic conditions, as well as variables of external economic conditions.