This paper analyses the spatial pattern of the agricultural yield data. Using the spatial statistics, it is possible to estimate some parameters of the semivariogram to study the problem of the systemic risk, which has great implications for the crop insurance program in Brazil. In particular, the “range” parameter was estimated. This parameter measures the distance, in which the spatial correlation tends to zero. For the empirical analysis, county yield data was used provided by IBGE, for soybean and corn, in the state of Paraná, through 1990 and 2002. The results showed that there is spatial dependence in every year analysed, going to zero in relatively long distances (in km).