Coverage assessment in an administrative census: A progress report on issues and methods reports on preliminary investigation into issues and methods for coverage assessment, if Statistics NZ moves to a Census of Population and Dwellings based primarily on administrative data.
The paper describes coverage issues already identified in New Zealand administrative sources. It also summarises the experience of countries that have recently moved to, or are considering a move to, a census that relies on population registers or other administrative sources to count their population.
The purpose of this paper is not to make final decisions about coverage assessment for a New Zealand administrative census, but rather to help guide decisions about where to direct more in-depth investigation.
Purpose
In March 2012 the New Zealand Government agreed to a Census Transformation strategy. This strategy has two strands:
- a focus in the short-to-medium term on modernising the current census model and creating efficiencies.
- a longer-term focus on investigating alternative ways of producing small area population and socio-demographic statistics, including the possibility of changing the census frequency to every 10 years, and exploring the feasibility of a census based on administrative data (Statistics NZ, 2012a).
The main emphasis of the longer-term strand of the Census Transformation programme is on the feasibility of producing census information from existing administrative data sources, as this aspect is the least understood. In addition, exploring the suitability of existing data sources to produce official statistics is part of our drive to be “an administrative data-first statistics office” (Statistics NZ, 2012b). The investigation takes a phased and iterative approach.
The first phase of the investigation (which this paper is part of) is designed to provide evidence that will inform decisions on the preferred direction for future development of the New Zealand census (see Statistics NZ, 2014 for an overview). The early focus is on developing an understanding of future census information requirements, and whether existing administrative sources can meet those requirements.
Our current understanding indicates that two approaches appear to offer feasible solutions for an administrative census if barriers can be overcome (Statistics NZ, 2014a).
The first approach is to create a national population register. This approach is the basis for successful international examples. However, New Zealand does not currently have a population register, and privacy, legal, and cost concerns mean that the creation of a register in the near future is unlikely. Therefore we assume there will be no New Zealand national population register in the next 10 to 20 years.
Another approach is to link multiple existing data sources. This shows promise as a statistical solution to produce population counts that would avoid the need to survey everyone. This approach is the current focus of investigations, including this paper.
An assessment of coverage issues is integral to the development of a census based on administrative data sources, and methods to measure and adjust for coverage errors need to be developed.