Agricultural Census 2022

To define the population frame for the complete enumeration of the Agricultural Census 2022, the lists of households, enterprises, and organizations engaged in agricultural activities were compiled, using the administrative and statistical sources, by the National Statistical Office in 2021. Depending on the sector different sources were used:

• For the livestock sector, the list of households and enterprises with livestock was obtained from the End-year Livestock Census 2021
• For the crop sector, the list of households and enterprises engaged in crop production was compiled from the End-year Crop Statistics 2021 (Form A-Agri-7). In addition, households, enterprises, and organizations identified in the livestock frame that had also carried out crop production activities in 2021 but were not included in the crop frame were added
• For the forestry sector, the list of enterprises and organizations engaged in forestry activities was compiled from the End-year Forestry Statistics 2021. Additional lists of forest user groups, cooperatives, and professional organizations were provided by the Ministry of Environment and Tourism and the Forest Research and Development Centre and were cross-checked against the existing statistical frame. Enterprises and organizations not included in the frame were verified by the provincial and capital city statistical departments, and those confirmed to be engaged in forestry activities in 2021 were incorporated into the census frame
• For fishery and hunting activities, the frame was developed using the list of enterprises and organizations included in the End-year Fishery and Hunting Statistics 2021, supplemented by lists of households engaged in these activities provided by the provincial and capital city statistical departments

The Agricultural Census 2022 used a combination of complete and sample enumeration. Complete enumeration was used for enterprises and organizations engaged in crop, forestry, fishery and hunting activities, as well as to the relevant units covered under the census scope. For the household livestock sector, sample enumeration was used.

Among the 246 302 households with livestock recorded in the End-year Livestock Census 2021, 188 605 were herder households. The sample size selected amounted to approximately 74 400 households, representing 30.2 percent of all households with livestock. The sample size was determined using the standard formula for estimating sample size using proportions. The calculation was based on a 95 percent confidence level. The key indicator used for the sample size calculation was the share of adult livestock losses out of the total number of livestock at the beginning of the year, based on the average of the last three years. This proportion was estimated at 0.03. The acceptable relative error was set at 4.1 percent, corresponding to an absolute error of approximately 0.0012.

The total sample size was allocated to provinces using proportional allocation. To support the production of estimates at the district level, the target population was stratified within each province. The stratification was based on the livestock size of households, using the number of the five main types of livestock owned by households. For each province, 5 to 6 sampling strata were formed. Within each stratum, sample households were selected using systematic sampling.

The statistical results were calculated using sampling weights. The base weight for each sampling stratum was calculated as the inverse of the selection probability, or as the ratio of the total number of households in the stratum to the number of sampled households in that stratum. The results from sampled households were then expanded to represent the total population of households with livestock in the corresponding stratum and administrative unit.

Although the initial sample allocation was made at the provincial level, the relatively large sample size and high sampling fraction, together with stratification by livestock size and systematic sampling, ensured that households with different livestock sizes were represented across districts. District-level estimates were produced by applying the sampling weights to the sampled households belonging to each district. As a result, the sample design was considered sufficient to produce estimates with an acceptable level of accuracy at the district level.