Irish Population Estimates from Administrative Data Sources, 2020

The Irish Population Estimates from Administrative Data Sources (IPEADS) research paper is from the CSO Frontier Series of outputs. It uses pseudonymised administrative data from public sector bodies and has been developed by the CSO as an experimental method of producing population statistics for April 2020. This report is the first time the CSO has published detailed population statistics derived solely from administrative data. It forms part of the development and modernisation of population and demography statistics for Ireland over the next decade. This modernisation of population and demography statistics will be required to meet new EU Regulations in this area which will require the provision of detailed small area population estimates on an annual basis. The publication is based on an experimental and developing methodology. It is intended to both illustrate the potential of administrative data to produce demographic statistics and highlight the challenges that arise. These are not the official population statistics and should be used with caution.

The administrative data landscape is constantly developing. This can bring opportunities and challenges. The methodology needs to be flexible to deal with changes in the data sources. It is important to note that when using administrative data to estimate the population, different methodologies will result in different estimated counts. For example, using a ‘signs of life’ approach will result in a higher count than attempting to apply a strict usual residence criterion. See the Methodology section for more details.

Data collection

Currently population statistics in Ireland are produced through the Census of Population and the Population and Migration Estimates. The primary source of data collection is the Census of Population which is generally carried out every 5 years. In between censuses, Annual Population and Migration Estimates are calculated by trending forward the previous Census of Population using births and deaths data from administrative records and migration estimates from the CSO Labour Force Survey.

In the Irish Population Estimates from Administrative Data Sources (IPEADS), all statistics are produced using data collected from administrative records only. These records were not initially created for measuring the population, but rather for service delivery and day to day operations of public bodies.  However, activity in administrative data can be a sign of presence in the State. This experimental work attempts to estimate the population of Ireland in April 2020 by measuring activity in the administrative data using a new methodology devised by the CSO.

At an international level, an increasing number of countries are moving towards greater use of administrative data in censuses, a move supported by the United Nations Economic Commission for Europe (UNECE) recommendations. This is evident from the UNECE Census Wiki, which compiles information on the 2020 round of censuses as reported by member countries, see Background Notes. UNECE reports note this move has been motivated by the benefits of using administrative data sources in census, including reduced cost, reduced burden on respondents, improved timeliness and greater frequency of results. It is becoming increasingly difficult to conduct traditional censuses with challenges experienced in recruiting field staff as well as establishing contact with householders.

The Census has been postponed twice in recent years due to unforeseen circumstances. In 2001, an outbreak of foot and mouth disease in Ireland delayed the census by a year. Presently, the census originally scheduled for 2021 has been delayed by a year due the global COVID-19 pandemic. Robust administrative population data available on an annual basis would offset some of the risk of an information deficit in the event of delays to the census and insulate users from the delays and postponements which have occurred with the traditional census model.

The approach is still experimental however and there are aspects of using administrative sources for counting the population that will require further development. These include the relatively limited range of variables available (e.g. commuting patterns, use of the Irish language, religion) from administrative sources compared with the Census of Population and the production of statistics at small geographic areas. The CSO access to some outstanding public sector data flows will help with both issues, as will the increased level of Eircodes being collected in the public sector with the move to online public administration of services. The CSO is committed to addressing these issues in partnership with Public Sector Bodies and will work towards consistently improving these estimates based on data improvements over the coming years. The key metrics in assessing these data improvements for the population count are the percentage of records with Eircodes, (currently estimated in the region of 60%) and the percentage of records coming from ‘real time’ sources (approximately 70%), where the activity recorded is from a recent time period. The CSO is committed to addressing these issues and working towards consistently improving these estimates over the coming years.

Defining the population

There are several ways in which a population can be counted. The most familiar approach is the one taken in the Census of Population, which records the ‘de facto’ population, or the number of persons present in the country on Census Night based upon completion of the census questionnaire. This count can include temporary visitors to the country and exclude persons who usually live in the country but are abroad on the night of the census.

This experimental publication attempts to count the ‘usually resident’ population of Ireland in April 2020. For someone to be usually resident in Ireland in April 2020, they will have lived in the country for a continuous period of at least 12 months including April 2020.

Usual residence is a widely used statistical concept, recognised by many international bodies, including the European Union.

In a survey or census it is possible to get the required information to determine usual residency by asking people specific questions. This is not possible using administrative data sources, but usual residence can be imputed by looking at the level of activity in different administrative data sources over a period of time. There are challenges with both methods of defining usual residence. For example in the census/survey based approach respondents may misinterpret the question, while in administratively sourced population estimates there will often be coverage issues in individual administrative datasets and definitional issues.

The IPEADS publication is part of a body of developmental work undertaken by the CSO in recent years to look at alternative methods to count the population. The methodology in this report aims to estimate an administrative count of the population based, as closely as possible, on usual residence by applying a specific set of rules around activity in the 24 months from Jan 2019-Dec 2020.

The estimates are produced by linking administrative records from various data sets that have been pseudonymised to maintain privacy. Rules are then applied to decide who should be included in or excluded from the usually resident population. In general, the rules apply as follows;

Include:

• persons (and, where identified, their current partners and/or children) active over a period of 12 months or more, including April 2020, in key real-time datasets, for example persons in receipt of weekly and monthly social welfare payments or in the PAYE system for employment and pensions.
• persons in annual datasets, for example enrolment in education or annual self-employed returns.

Exclude:

• persons not recorded as active in administrative data between April and December 2020.
• persons active for a period of less than 12 months around April 2020.
• persons born after April 2020.
• persons who died before April 2020.

It is important to note that when using administrative data to estimate the population, different rules to the above could be applied. This would result in different estimated counts. For example, applying a rule that any person with any activity recorded in administrative data in 2019 or 2020 would be considered usually resident would result in a higher population count than is recorded in this publication. Conversely, a rule counting only persons with recorded activity in administrative data strictly within 12 months around April 2020 would result in a smaller population count. See also Methodology section.

Publication format

This frontier publication provides estimates of the usually resident Irish population in April 2020 and is divided into the following parts;

• The population estimates section outlining statistics such as;
• the breakdown of the population by age, sex, nationality and other variables of interest.
• geographical results including a breakdown of the population across counties and electoral divisions.
• A methodology section outlining how the population was derived for this publication.
• The background notes section giving further information about the different administrative datasets that were used and further information about the privacy measures undertaken.

Key findings

• Based on this experimental methodology the population of Ireland in April 2020 was estimated to be 5.2 million. The age structure of the population reflects the impact of changing birth rate and migration patterns over recent decades.
• Statistics on age, sex, marital status, nationality, principal economic status and NACE sector can be produced and broken down by county.
• There is the potential to produce population statistics for electoral divisions and other geographical boundaries such as local electoral areas and small area data. Improved coverage of Eircodes on administrative datasets is key to producing these outputs and will expand the provision of small area data.

Note that this experimental publication is focused on the production of population statistics. The CSO will continue to expand the methodology developed for this project and aim towards developing the following areas;

• Produce population estimates from administrative data on an annual basis.
• Improve timeliness by producing the population estimates within one year of the reference period.
• Develop population estimates and related statistics for geographical boundaries such as local electoral areas, electoral divisions and other geographical boundaries.
• Create a longitudinal series with annual linked datasets. This will provide a more stable population base over time.
• Develop collection of more census type variables, for example housing data looking at the total housing stock, vacancy or central heating, internet connectivity and other attributes linked to dwellings.
• Carry out a detailed comparison of the Irish population estimates from administrative data sources with the 2022 census of population results.
• Work towards a combined census consisting of administrative data and a large-scale survey in 2031.
• Explore how development in administrative data and working patterns because of the COVID-19 pandemic affect the population estimates.

This report is a further example of the policy-relevant research projects the CSO is developing as part of its leadership role in the Irish Statistical System. Our goal is to maximise the variety and volume of data available to provide high quality information to the Government, businesses and citizens, through the development of a National Data Infrastructure (NDI).

The NDI plays an integral part in facilitating the CSO to develop new and improved statistical products for the benefit of citizens and policymakers. The core concept of the NDI involves the collection, maintenance and storage on all public-sector data holdings, of the associated PPSN, Eircode and Unique Business Identifier (UBI) to be developed whenever they are relevant to Public Sector Body transactions with customers. This supports the development of targeted policy interventions.

Under the auspices of the Statistics Act 1993, and in compliance with all relevant data protection legislation, the CSO is in a unique position to gather and link administrative data sources held by Government departments and agencies and evaluate their potential for statistical use.

The demand for data and insight into Irish society continues to grow unabated. The growth is not just apparent in terms of the broad range of themes (e.g. globalisation, productivity and well-being) that Official Statisticians are being asked to provide information on, but also in relation to the level of detail being required in the analysis (e.g. socio-demographic variables). Clearly the range and depth of demand for data cannot be met from survey data alone. Analysis of new data sources is imperative, including administrative records held by Public Sector Bodies.

The CSO Statement of Strategy and National Statistics Board Strategy highlight the importance of the adoption and implementation of the NDI to support evidence-based policy making. Significant progress has been made in recent years but perhaps now is the time to accelerate the implementation of the NDI across all Civil and Public Service data holdings. This publication highlights the real potential that exists in these data holdings to support improved evidence-based policy making in key areas such as housing, healthcare, education.