Abstract

The paper will overview details of the sampling methodology practices for the survey of agricultural holdings – the important source of official agricultural statistics of Georgia conducted permanently by the National Statistics of Georgia (Geostat).

A good elaborated sampling methodology is very important within statistical surveys and it reflects on the whole statistical processes, from data collection to processing. Therefore, a sample planned on a wrong manner directly cause errors in terms of final statistical data production. Thus, errors in sampling methodology can decrease entire trust towards the statistical information throughout data stakeholders.

The sampling design crucial regarding various statistical assessments and estimations, calculation statistical errors. This last one is significantly dependent on the standard error of estimation, whose calculation formula essentially provides sample design.

A general source of sampling of the survey of agricultural holdings of Georgia is the dataset of the agricultural census conducted in 2014. The sampling size for the survey of agricultural holdings covers about 12,000 holding. Noteworthy, that total number of holdings in sampling frame amounts 640,000.

A two stage stratified sample is used for the survey of agricultural holdings, where the strata is defining by three following factors:

1. A geographical region – it implies a region where the holding operates. Noteworthy, if the holding operates in more than one region, as the “statistical region” of a holding would be considered accordingly to FAO’s methodology.
2. Legal status – this is another strata actively used in sampling of the survey of agricultural holdings of Georgia and allows to distinguish family and non-family holding (enterprise).
3. Size of holding – in order to identify size of the holding, an aggregated indicator agricultural index is used, which is calculated throughout all agricultural assets of the holding (agricultural land (sown/arable land, permanent yards, meadows, and pastures, greenhouses) and livestock in one unit. convention carried out on the basis of the economical comparison and for benchmark unit was considered one hectare sown of the maize (the conventional factor of land that doesn’t give economical input in direct form (such as uncultivated arable land, meadows and pastures, etc.) evaluated by minimal potential or some other methods).

It should be noted, that size of the holding is divided into 4 groups - (small, medium, large, and extra-large holdings) in each region. The borders of stratums are determined individually by the method of square roots. In small and medium holdings there wasn’t revealed a necessity of separating family holdings and enterprises due to the nonexistence/few numbers of such size enterprises, the similarity of scale, or non-significant statistical difference between probabilistic distributions; there was no necessity in extra-large holdings also because this type of holdings appearing in the sample with probability equals to 1. Therefore, only the stratum of large holdings is split by family holdings and enterprises stratum. As a result, we have 55 stratums (11 regions X 5 size/legal status). For redistributing of sample size by the stratums was used von Neumann’s allocation principle (redistributing of whole sample size by the proportion of multiplication of the number of holdings in stratum and standard deviation in stratum (by the agricultural index)).

The paper will be in touch with the issues related to stratification and sampling. the paper will help young experts and researchers to get additional information on experience adopted in Georgia, through cooperation with international partners.