Abstract
In 2022 the third edition of Anders and Britt Wallgren’s book on register-based methods was published by Wiley. It has been substantially restructured since the second edition, and is now shorter by some forty pages, but a large proportion of the material is common between the editions. The latest edition is focused more on the building of a national statistical system on the basis of register sources. It includes two chapters and several further discussions of the business register and the national accounts which were previously not strongly represented.
This edition still talks about a system based on administrative data as a new approach requiring a new paradigm and new methods. Since this edition appeared almost twenty years after the first (Swedish language) version, and the system has been in development even longer, I suspect that we have moved to the point when this approach is no longer new. But there are many countries where steps are being taken either to move to an administrative data based system, or to make substantially greater use of administrative data as the basis for official statistics, and of course the ideas may be new in these contexts. Indeed these are called “new register countries” in the book, and many of the structural changes are to make information about this transition more explicit.
Several new examples have been added, including several from the authors’ experiences teaching administrative data methods in the Caribbean and Latin America, though the core of the examples originates from their work with the Swedish statistical system.
The first chapter sets the scene, explaining the difference between register surveys and the census plus sample survey approach (implicitly to population estimation) used elsewhere. It sharpens the focus on the statistical system compared with the second edition, and already introduces the four base registers as a framework for such a system. It mentions big data, but argues that this is a different concept than administrative data, so it is not treated in this book.
The second chapter deals with the transition to an administrative data system, and the challenges that it presents, from the need for legislation to the requirement for all the actors in the system to work together with common objectives. The third chapter introduces some of the basic characteristics of administrative data, including some examples contrasting them with data collected through questionnaires.
Then there are three chapters dealing with methodological issues associated with “Building the System,” covering record linkage, quality assessment, and data editing, all of which are essential components for assessing and combining administrative data. Record linkage methodology has advanced considerably in recent years, and this section has been expanded, giving some examples and identifying some causes of linkage errors. I would have welcomed a short discussion of how to deal with these kinds of errors in the system. The quality assessment chapter is wider than assessment – it includes inventorying available administrative sources and a more active approach where the statistical system takes the initiative to improve administrative data systems and the administrative data landscape, to help improve the statistics. It gives examples where this is happening. It also deals with more traditional concepts of quality measurement, for which there is less theory than for sample surveys, but where there are some helpful approaches. Some measures can be obtained through surveys to measure quality of the registers constructed from the administrative data. Data editing runs from editing a single source to the topic of consistency editing, which is a unique set of processes for linked administrative systems. These are again supported by helpful case studies.
Then three chapters deal with the population register, the first about its construction, and then chapters about estimation methods and coverage errors. Chapter 7 deals with the challenges faced in constructing a population register for the first time, and includes only a short retrospective of issues for countries which already have such a register. It includes the “signs of life” approach to identifying which people should actually be present, which has become a key concept for administrative data censuses in recent years. The estimation chapter gives a brief overview of methods for administrative data, contrasted with surveys. It describes calibration approaches and includes some new examples from the Swedish LFS. The coverage errors chapter has population and business register examples (the latter before the construction of the business register is described).
Chapter 10 deals with the construction of the business register and how it can be a driver for consistency of estimation within the national accounts. This is a substantially expanded topic from the second edition, and includes some useful information on the way different business units contribute to different measurements (though the authors could perhaps have mentioned “the unit problem” (van Delden et al. 2018)). It is also furnished with some nice examples of the discrepancies which can arise through using different administrative sources, and the ways in which these can be handled. The following chapter deals with estimation with the business register, particularly covering multi-value variables and one particularly important application, to industrial classification.
The final chapter is called “Censuses, Sample Surveys and Register Surveys – Conclusions,” but it deals with rather more than conclusions—it provides some history of the administrative population census, and a summary of some of the concepts which have been introduced in previous chapters. It also adds the idea of register survey design.
The index is longer than in the second edition, which is helpful in locating topics in the new structure if you are used to the previous edition.
This is a very valuable book, covering a subject for which there is relatively little literature. The new edition is brought to life with many examples including some new ones, from the authors’ experiences, both of working within the statistical system and of teaching these methods to others. The new structure is broadly logical, though there are some places where components feel a little out of place. The additional material on the business register and how it feeds into the national accounts rounds out the treatment, which was more focused on the population register and associated register surveys in the second edition. The second edition had some more details about the ways in which registers and the system are constructed, and I will still find my copy useful. But the latest edition promotes the role of the statistical office in developing the administrative data system, and this change to the framing of the book is an important reflection of the development and maturity of the administrative data based statistical system.
