The Irish PECADO project: Population Estimates Compiled from Administrative Data Only

Dunne, John (2020) The Irish PECADO project: Population Estimates Compiled from Administrative Data Only. University of Southampton, Doctoral Thesis, 200pp.

Record type: Thesis (Doctoral)

Abstract

This thesis proposes a new system of Population Estimates Compiled from Administrative Data Only (PECADO) for Ireland.

Ireland does not have a Central Population Register (CPR) upon which to develop population estimates in the manner of the Scandinavian and Dutch models. Ireland does have a strong system of Person Identification Numbers (PIN) that are used across public administration systems when a person interacts with public services. To the knowledge of the author no statistical agency in the absence of a Population register has yet compiled population estimates using administrative data only.

The PECADO system of population estimates takes as its starting point the compilation of a Statistical Population Dataset (SPD) from administrative data sources using a signs of life (SoL) approach. The SoL approach only includes persons in the SPD where there is strong evidence that a person is alive and living in the state for a significant part of the reference year. The SPD is compiled with respect to a reference year. The SoL approach does not accept a person's registration on an administrative system as sufficient evidence for including that person in the SPD. The SPD counts are then adjusted for under-coverage using an adaptation of Dual System Estimation (DSE) methods. The second list or list B that is used in the PECADO DSE setup also comes from an administrative data source not previously used in the compilation of the SPD.

The thesis considers the traditional DSE approach in the context of a Census under-coverage survey (UCS) (Wolter, 1986) and presents an alternative formulation of DSE methods that allows a relaxing of the strict assumptions associated with the traditional approach. This alternative formulation now facilitates DSE methods being applied in a much broader set of circumstances, in particular, where one list is derived from administrative data sources and the second list acts as the capture list where each person in the population has an equal probability of being caught (homogeneous capture assumption). The thesis then proposes an extension to the DSE methods, Trimmed Dual System Estimation (TDSE), that provides a tool to allow for the evaluation of suspect parts of the SPD for erroneous records. The thesis also considers the situation where the homogeneous capture assumption is weak and discovers that in certain situations this assumption may also be relaxed without introducing bias to the population estimate. We label these set of tools and methods the PECADO toolkit.

The thesis presents and evaluates a set of population estimates using the PECADO toolkit. In particular, the thesis shows this system of population estimates to be plausible. The plausibility of the estimates is demonstrated by using the tools in the PECADO toolkit to provide reassurance the SPD does not contain erroneous records and to provide some reassurance that the second list compiled from administrative data sources does not introduce bias. The population estimates show some differences when compared with the Social Census counts. The conceptual differences can be reconciled when migration is factored in but the differences between the estimates are too large for this to be the sole explanation.

The underlying DSE methodology is also considered with respect to using an administrative data list in place of a second eld operation to adjust traditional Census counts for under-coverage. We consider the 2016 Census. Ireland to date has not conducted a Census coverage survey and has relied on the diligence and motivation of the Census field force to ensure everybody is counted. Ireland plans to conduct a traditional UCS with a field operation as part of Census 2021. If an administrative list can be used instead of a second field survey then this approach will save significant time and money while also simplifying the application of underlying DSE methodologies.

The PECADO system of population estimates is then further extended to estimate population flows, reusing the same administrative data sources as were used to compile population (stock) estimates. The resulting estimates of population flows (inflows= births + immigration, outflows = deaths + emigration) readily reconcile with the PECADO population (stock) estimates as this extension incorporates a coherent demographic accounting framework.

The thesis concludes with some consideration of possible next steps that could be taken to disaggregate population estimates by detailed geography, how to incorporate attributes and how to build household units with a view to being able to compile Census like estimates on an annual basis.