Non-traditional data sources can improve population and housing statistics

Official statistics depend on traditional methods and primary data sources such as censuses, surveys and administrative data. These ensure consistency and reliability but are generally costly and not very flexible or timely to capture emerging trends or phenomena. The increasing availability of non-traditional data sources offers new opportunities to improve the production and quality of statistics. In the specific context of population and housing statistics, non-traditional data hold potential for filling gaps in official statistics, overcome measurement errors, and provide higher spatial and temporal granularity, enabling more accurate, timely and detailed analyses. This report documents the results of a 2-year project carried by the Joint Research Centre in collaboration with Eurostat entitled ‘Testing the feasibility of using Non-traditional data sources to IMprove Population and Housing Statistics’ (NIMPHS). The main objective was to support Eurostat in exploring the potential and limitations of non-traditional sources such as remote sensing and mobile phone data to improve the quality, detail and/or update of population and housing statistics. In doing so, the JRC also produced and released two new pan-European flagship products by combining existing census statistics and fine-scale data derived from remote sensing: a population grid map at 100 m resolution and a housing grid at 1 km resolution. In addition, the project developed and tested a prototype approach and tool to ‘nowcast’ population grids, and thus improve the timeliness of gridded population data. The report discusses and draws conclusions and implications of the work carried throughout the project. Non-traditional data sources hold substantial potential to improve certain dimensions of pan-European population and housing statistics, namely by increasing their detail, timeliness and overall quality, namely through contributions to quality assurance and control. However, such data sources come with challenges, preventing their direct translation into official statistics. Their use requires good knowledge about the details of the data to devise well-thought data cleaning, processing, and fusion protocols.