Črt Šuštar, Statistical Office of the Republic of Slovenia

Nowadays, official statistics in Slovenia relies heavily on administrative data. Despite reliability and extensive coverage of official registers, not all information is explicitly captured in individual data sources. In those cases, either survey questions or data modelling has to be utilized.
In the scope of the Modernization of Agricultural Statistics project, modelling processes were developed in order to extract desired information by combining multiple data sources. Besides ‘traditional’ statistical modelling methods, also geospatial modelling was applied. This enabled a more thorough exploitation of the data sources by utilizing the information on the spatial dimension.
As a first pilot, the ownership structure of the agricultural land was analysed by combining two geospatial and three tabular administrative data sources. Results were proved to be reliable by validation through historical data. Therefore, the process will be utilized for data collection in the Agricultural Census 2020. The same processing logic will also be applied for obtaining data on agricultural land structure, irrigable land, rough grazing, crop rotation and market garden/open field differentiation.
The elementary units for modelling are agricultural plots, instead of agricultural holdings, thus the granularity of the obtained data is much finer. This opens possibilities for various data aggregations and further modelling. The developed models enable us full capture of the situation for a reference date without creating any respondent burden. The same principle is usable in other areas of statistics.