Methodological approach for determining representative sample of Bulgarian farms

Minka Chopeva

Abstract: The aim is to present the methodology for determining a representative sample of farms and the results of its application. The goal is conditioned by the fact that much of the farm research is based on information from sample surveys. A method of the regional division combined with models of two-stage cluster sample is used. In determining the number and composition of areas and farms are respected. The requirements to fall areas and farms in the sample, which reproduce in adequate level structure of agricultural holdings in their specialization are taken into account. There are an estimated two dispersions, one, called the inter-cluster which represents the magnitude of the dispersion in each planning region, and the second relates to the dispersion inside the districts. It is done in order to achieve the necessary consistency between the farms structure in general aggregation and their structure in the sample. Additionally two indicators to verify the representativeness of the sample are used. The first reflects the degree of correspondence between the average number of farms in one nest of the sample and the average number of farms in one area in the general aggregation. The second indicator measures the degree of correspondence between the variation in chosen nest of the sample and the total variation in the general aggregation. Considering the obtained values of inter-cluster and intra-cluster dispersion, the following rule have used. First, from each planning region is selected this district that most closely approximates the average of farms structure in the region. Additionally selected those districts that have the greatest deviation from the average structure in the relevant planning region. Thus it is ensure that the sample will fall districts sufficient to cover production diversity in the region and will reproduce the structure of farms. The scope of the sample included farms from 20 districts in the country. Districts whit missing farms in the sample are those whose specialization structure is very similar to the structure of the neighboring districts. The maximum permissible error at guarantee multiplier (z = 1.96) is equal to 4.74%. Therefore, it can be argued with guarantee 95% probability that the actual values of the evaluated parameters from the questionnaire information will move within a confidence interval of ±4.27%.

Keywords: districts; farms; maximum error; representativeness; sample

Date published: 2017-10-09