The Control of Non-Sampling Errors on Linked Data: an Application on Population Census

Error management represents one of the main aspects of quality control in the statistical data production process. For more than ten years quality improvement has been a strategic objective for the Italian Statistical System (SISTAN). Evaluation and quality control still cannot be considered the final elements of the production process but strictly belong to the survey design and hence the planning phase. Record-linkage represents one of the elements that could best assure advantages from the informative patrimony accumulated by data producers, both in the way of a greater supply of joint statistical information on the same population, in the building of statistical informative systems or integrated (or federated) archives, and in the way of the data quality improvement of a survey through information obtained from other surveys or administrative archives. However, counter to the potential advantages offered, it is necessary to take into account the complexity and the further uncertainty that record linkage can also introduce. With this aim in mind, the present paper will attempt to verify how it is possible to discern between response errors and under-coverage errors in the Italian Population Census of 2001, through a post enumeration survey and in so doing, apply a record linkage procedure that could reduce these errors but give rise to other sources of error. These different sources of error are defined in in the paper, as well as their interaction with linkage errors. It is also presented an application regarding the estimation of some characteristics of the foreign resident population.