The paper is dedicated to the relationship of data aggregation level and yield variability. For that purpose yields of the major crop plants in Poland are analysed i.e.: winter wheat, triticale, rye, barley, oat, mixed cereals, rape and sugar beet. The research are based on data from Polish FADN from years 2004 – 2009. The samples’ size ranged from 531 to 2893, depending on the plant crop. In the paper six levels of data aggregation are examined, that is: farm, district, powiat1, voivodship, region and country. It was found out that the degree of yield variability reduction (observed with data aggregation) is crop specific. Nevertheless, the relationship between aggregation level and yield variability can be approximated by the same formula for all of the investigated crop plants: ... , where MUA is the average production area in the administrative unite.