Assessment of the possibility of using proportional stratified sampling for estimating resources of agricultural farms.

The article is dedicated to studying the possibility of using proportional stratified sampling as a method of survey observation in the agricultural census aimed at obtaining representative data on the resources of personal subsidiary farms. The purpose of the work is to test and evaluate the effectiveness of this approach to reduce data collection costs and improve the quality of statistical estimation of resources of agricultural producers in pilot municipal districts. Additionally, the article addresses issues related to the construction of stratification variables, calculation of sample sizes by strata, weighting procedures, and correction of biases, as well as criteria for accuracy and robustness of estimates. Field observation schemes, quality control algorithms, and approaches to assessing budget savings are presented. The analysis includes validation of the obtained estimates using benchmarking, bootstrap resampling, and comparison with complete local censuses. The study used the method of proportional stratified sampling with stratification by localities of municipal districts and random selection within the strata, as well as the calculation of representativeness errors by comparing sample and population data on crop areas and livestock numbers. The novelty of the research lies in the fact that, for the first time using real microdata from the Agricultural Census of 2016, practical testing of the sampling approach has been conducted for the preparation of the All-Russian Agricultural Census 2027. Methodological and operational limitations of proportional stratified sampling were identified when dealing with scarce and spatially unevenly distributed resources. Satisfactory results were obtained only for widespread and evenly represented indicators (such as main crop areas and large livestock), while for rare crops, niche directions, and small groups of animals, the sampling representativeness proved insufficient even with post-stratification and model-based overestimation of small areas. Considering the requirements for completeness and comparability of data for policy monitoring, it was recognized that a complete method is preferable for the All-Russian Agricultural Census 2027, while sampling procedures are reasonably used as a supplementary tool for calibration, post-census checks, and quality control.