Abstract

New data sources, denoted as Big Data (BD), have emerged and are the result of interactions with digital technologies by citizens and business units and the increasing capability of these technologies to provide digital trails. The BD sources could represent an effective reply to the declining response rates and the rising costs of conducting surveys since they can provide further useful information in the statistical process. Their use poses new challenges according to a paradigm shift: from designed data with a probabilistic sample to data-oriented or data-driven statistics with a non-probabilistic sample. The paper focus on the data integration approaches combining multiple sources (surveys with probabilistic samples, administrative data and BD sources) to improve the accuracy of Official Statistics. Its main purpose is to define a statistical framework to make valid inferences that is congenial to the data production process of a National Statistical Office and is appropriate for the official statistic goals.