Pushing the boundaries for automated data reconciliation in official statistics

Cover Pushing the Boundaries for Automated data Reconciliation in Official Statistics, Jacco Daalmans
© CBS
Dissertation on methods to resolve inconsistencies in statistical output at aggregate level.
National Statistic Institutes, like Statistics Netherlands, have to publish reliable and coherent statistical information. To meet this requirement, different results on the same phenomenon should be the same. If different outputs are based on different sources and processing methods, consistency does not happen naturally. Macro integration is a process to arrive at consistent results. The main idea of most macro integration methods is to ‘minimally’ adjust preliminary output to satisfy a set of predefined consistency constraints. This thesis extends currently available macro integration methods, further explores properties of existing methods and examines new application possibilities of macro integration methods.

Chapter 2 proposes an extensions of a popularly applied Denton method that combines several features of other macro integration methods. Chapter 3 argues that an often advocated “Growth rate preservation method” has certain practical drawbacks that are not often mentioned in the literature. Chapter 4 explains that a well-known Denton method does not perform well in a “sequential estimation” setting and proposes solutions to this problem. Chapter 5 examines the use of a macro integration approach for the estimation of Dutch census tables. Chapter 6 considers the problem of the simplification of a set of consistency rules, which is important for any application to a large problem.

Daalmans, J.A. (2019). Pushing the boundaries for automated data reconciliation in official statistics. Dissertation, Tilburg Universitiy.