The pitfalls and benefits of using administrative data for internal migration research: An evaluation of Australia’s Person Level Integrated Data Asset (PLIDA)

Aude Bernard; Jing Wu; Tom Wilson; Neil Argent; Tomasz Zając; Anthony Kimpton

doi:10.4054/DemRes.2024.51.22

Volume 51 - Article 22 | Pages 687–722

The pitfalls and benefits of using administrative data for internal migration research: An evaluation of Australia’s Person Level Integrated Data Asset (PLIDA)

By Aude Bernard, Jing Wu, Tom Wilson, Neil Argent, Tomasz Zając, Anthony Kimpton

Date received:

25 Jun 2024

Date published:

1 Oct 2024

Word count:

7704

Keywords:

administrative data, Australia, internal migration

DOI:

10.4054/DemRes.2024.51.22

Download PDF
Submit a Response Letter

Abstract

Background: To enhance its data capability, Australia recently set up a longitudinal administrative micro-dataset, the Person Level Integrated Data Asset (PLIDA).

Objective: To ensure that users in both scholarly and applied settings understand how PLIDA can be reliably used, we assess its Combined Location Module, which provides place of residence by combining three administrative datasets since 2006.

Methods: Using descriptive statistics and regression analysis, we compare the population coverage of PLIDA to the census, estimate the incidence of missing values at various spatial scales, quantify spatial mismatch between PLIDA and the census, and compare the intensity, selectivity, and spatial patterns of internal migration between the two datasets.

Results: The PLIDA population is higher than the census population, but very remote populations and recently arrived and temporary migrants are under-represented. We uncover a high mismatch rate between PLIDA and the census for small spatial units, particularly among highly mobile groups. As a result, PLIDA overestimates the level of internal migration and slightly distorts migration age patterns at young ages. Migration patterns are broadly comparable at large spatial scales, except for rapidly growing peri-urban regions, inner cities with short-term residents, and regions with a seasonal or temporary workforce.

Conclusions: PLIDA can be used in some applied settings outside demography if the spatial scale of analysis is aggregated, the analysis is restricted to census respondents, or the analysis is limited to well-represented groups. However, in its current form PLIDA is not suitable for demographic applications such as internal migration estimates, and we discourage its use for population projections.

Contribution: Researchers need to be aware of the pitfalls of administrative data to knowingly decide on appropriate use. We recommend researchers to keep abreast of developments by the Australia Bureau of Statistics, which is continuously improving PLIDA.