2023 Post-enumeration Survey: Final linking methodology details the final 2023 Post-enumeration Survey linking design and changes made to the design described in 2023 Post-enumeration Survey: Linking design. It also provides the final 2023 Post-enumeration Survey linking results.
Summary
The 2023 Census Coverage project aims to measure the coverage of the 2023 Census which is a combined census by design, incorporating both census responses and admin enumerations.
The Post-enumeration Survey (PES) is a household survey undertaken shortly after the census to evaluate the completeness of census coverage. A key part of the coverage estimation used at Stats NZ is linking PES responses to the census records (both responses and administrative data enumerations) to find out who was missed by the census, counted more than once, or counted in error. The outputs of this linking process inform the rest of the coverage estimation process.
One of the assumptions that underpins the coverage estimation is perfect linking - that every PES record is linked to every census record that refers to the same person. However, linking comes with potential linkage error which can result in two records being incorrectly linked or two records that refer to the same person not being linked.
The 2023 PES linking design aimed to build on the lessons learnt through the 2018 linking process (Stats NZ, 2023b). This included use of triangular linking - using administrative data to evaluate suggested links between PES and census. The 2023 design also had a much more comprehensive clerical linking and quality assurance process to minimise biases that may occur through automated linking. This paper outlines improvements made to this initial design during the linking process.
The 2023 PES linking process was successful in achieving all perfect linking metrics. The results outlined in this paper also outline which parts of the process were most successful at linking different demographics and different methods of being counted in the census to help with future linking design decisions.
The paper is broken into the following sections:
- Input data used in linking describes the datasets used in the PES linking processes.
- Assumptions and methodology for linkage counts and percentages gives details of assumptions that are used to find values such as percentage of true links.
- Automated linking methodology describes the original plan for PES-census and PES-IDI spine automated linking work, followed by updates made and their impacts. A more detailed analysis on the impact of the updates of the PES-census linking work are in Impact of changes to pass design and Changes to pass 2.
- Original clerical linking design and results describes the original clerical linking process and the results of the first round of quality assurance.
- Changes and improvements to clerical linking design describes changes made after the first round of clerical linking work to improve methodology and ensure the quality assurance targets were met. More details on clerical linking results are in Appendix 5: Clerical linking and linking mitigation work.
- Final linking process outcomes and results provides the over-all results and link rates from the full linking process.
ISBN 978-1-991307-54-5
Enquiries
Susan Jowett
[email protected]
