Chair: Claudia Junker
Room: S3B Sukiennice
Time: 11:30 - 13:00
Date: 27 June
|Title: <<< Record linkage methods for Admin Data: Portuguese Census Transformation Program >>>
From 2014 to 2016 Statistics Portugal (SP) has been studying the usability of available administrative data for census purposes. Unlike other countries that have already made the transition to register-based or combined census models, in Portugal there is neither a central population register nor a unique personal identification number. SP created a methodological framework to build a Statistical Population Dataset (SPD), integrating linked registers in a “potential” resident database, then applying a ‘signs of life methodology’ to estimate residents . For 2015, the population estimated from the SPD was 10,434,161 persons, with a deviation of 0.9% to the official resident population estimates for the same year. Results are promising at a national level, but still, there are multiple hurdles to the creation of this dataset that don´t particularly benefit the results at a detailed geographical level: records have inconsistencies and errors due to manually inserted data, and the National Data Protection Authority imposed anonymization criteria on the datasets, restricting access to the full name and address of the persons in registers. Exact comparison methods performed by SP left out many potential matches (roughly more than 5% for most sources). With the goal of identifying the highest number of linked pairs of records, a cooperation a between SP and the University proposed an alternative linkage model , based in a logistic regression, which added thousands of new pairs of linked registers to those found by SP. The precision of the method is about 99% on a large set of linked records used as gold standard.
Statistics Portugal. 2016. Metodologia atualização Base População Residente – Construção BPR 2015. 2021 Census Unit.
Silva, R., Sampaio, L., Calado, P., Silva, M., Delgado, A. ‘Matching administrative data for census purposes’. In Proceedings of the Data Science, Statistics and Visualization Conference, Lisbon, Portugal, July 2017.
|Marina Pérez Julián|
|Title: <<< Assessing quality in a register-based census >>>
Spain is going to conduct for the first time a Census which will be largely based on registers. It will be built around the Population Register and will integrate administrative data from many sources. In order to help users to understand the production process and to better interpret the results, for each variable, another categorical variable measuring the quality of the information will be disseminated. These variables provide a powerful way for assessing quality in the census data across columns (variables) and rows (people and households). As the process for elaborating the census 2021 is ongoing, the intermediate results of this analysis is also helpful for focusing our efforts. In this paper we present some preliminary results based on the first general test of integration of sources we have done so far, the so-called pre-census file 2016.
|Title: <<< Quality as a priority of 2021 Population census? >>>
In the last 2010 round of censuses, we have witnessed historic transformations of unprecedented scope. The main motivation for changes in census methods was cost reduction. The arguments also indicated willingness to improve quality of estimates, timeliness and form of data dissemination due to introduction of modern technologies (UN, 2013). The rolling census methodology was first applied in France. In the United States, the American Community Survey replaced traditional long form. Alternative census methods using administrative registers increased interest. This approach stimulated development and application of modern methods of imputation, statistical data integration, calibration and small area estimation. The process of changes is not finished and it is difficult to predict its consequences. The discussion is lively, and some findings suggest 'the end of censuses' (Baffour, King, & Valente, 2013; Coleman, 2013; Poulain & Herm, 2013; Valente, 2010). However, survey conducted for 1985 - 2014 indicated that the number of countries carrying out census increased, and methodology transformation concerns Europe (Kukutai, Thompson, & McMillan, 2015). The aim of this study is to present changes in methods of conducting censuses from the perspective of quality assessment. It attempts to show the scale and diversity of proposed solutions indicating their possibilities and limitations. Assessing quality, references were made to significance of censuses, their role for state administration and international comparisons (Baffour et al., 2013, p. 409). In accordance with Brackstone’s definition (1999, 2003), all aspects of statistical survey reflecting usefulness to users were analysed. Similar, utility concept of quality was developed for official statistics and international organizations (Eurostat, UN, IMF, OECD), though components and classifications differ (Vries, 2002). The study took into account following six dimensions: relevance, accuracy, timeliness and punctuality, accessibility and transparency, comparability and coherence (Eurostat, 2003). The analysis concerns Polish experiences and perspectives, but includes references to other countries.
|Title: <<< Selected Innovation Elements in Census 2021 >>>
The next population and housing censuses (the Census) taking place in Slovakia in 2021 is, in a way, unique. The change of concept is its essential feature. It is a transition from traditional population and housing censuses to combined ones based on the integration of data from registers and administrative data sources and data obtained from population. The implementation of the new concept covers several elements of innovation, including full electronic data collection, systematic evaluation of quality of administrative data sources used for statistical purposes, coordination of the Census preparation via the National Action Plan and data integration. The paper is focused on the presentation of those innovations that are specific to that change and are substantial to the preparation of the Census, which is labelled, in conditions of Slovakia, by the term “integrated Census”.