A Longitudinal Error Framework to Support the Design and Use of Integrated Datasets
This chapter aims to contribute to the evolving literature on the use of administrative data for statistical purposes. It describes two frameworks for understanding statistical error. The first applies to single administrative sources that may be longitudinal in nature. The second applies to multiple sources that have been longitudinally linked. These frameworks support statistical design and the evaluation of administrative data quality. They may also support the construction of population estimates based on administrative data and possibly survey data in countries that do not have population registers. The frameworks were applied in the transformation of the population and migration statistics system at the UK Office for National Statistics. The chapter draws a distinction between operational and statistical data quality, demonstrating the tensions that can exist between these distinct uses of the same information.