1. Introduction

The census of population and housing is the single largest statistical exercise undertaken by government in Northern Ireland, it is also the most important source of information on the size and nature of the population. Due to the prominence and importance of this data there is an intrinsic need for accuracy, thus every effort is made to count everyone. The census is the only source which provides a detailed picture of the entire population, and is unique because it counts everyone at the same time asking the same questions of all. Subsequently, the information from the census is used widely and underpins national and local policy making.

The most recent census was held on 21 March 2021. The purpose of this document is to provide background information on the processes that underpin the census outputs. Through outlining these processes, users can gain an understanding of the underlying quality of census outputs. In many instances, links are provided to papers that describe aspects of the process in greater detail; these papers will be expanded and developed as further output is published.

While Census Office strive for precision, no census can be perfect – some people are inevitably missed and some people do not provide information for every question. Consequently the Northern Ireland Statistics and Research Agency (NISRA) uses peer reviewed internationally developed statistical techniques to make allowance for any missing people or information. NISRA’s aim is to ensure that census outputs reflect the whole population of Northern Ireland, such that government, health and education authorities, and many other organisations can target resources more effectively and plan services.

Census Office had six key strategic objectives for Census 2021, the first of which was to provide high quality, value-for-money, fit-for-purpose statistics that meet user needs, and which are consistent, comparable and accessible across the UK.

Details on the quality measures undertaken to assure the census data are covered in some detail in the quality assurance report being published alongside this document.

2. The census process

2.1 General planning

Carrying out a census is a large and complex task, and planning for a census takes a considerable amount of time. Following on from the 2011 Census, NISRA, along with other UK Census Offices, looked at other ways to collect and use census information. In 2014, NISRA recommended to the Minister of Finance that a modernised census in 2021 was the best way to meet the need for the data only a census collects.

This was accepted and NISRA published The Future Provision of Census of Population Information for Northern Ireland in October 2014. As part of this publication NISRA asked for any comments or feedback. This was the first engagement on Census 2021.

A major part of the planning for a census is consulting and engaging with stakeholders (including the public and users of census data). The formal planning for Census 2021 started in September 2015 with a topic consultation asking users what topics and questions they might like to see in the census. This formal consultation included a series of roadshows. The details of this and all subsequent consultations, can be seen on the consultation section of the NISRA website. Consultation and engagement continued throughout the census process and at the time of writing, the most recent consultation outlined plans for the content and dissemination of census outputs. In addition to events such as public meetings, NISRA has consulted regularly with key advisory groups such as the Demographic Statistics Advisory Group (DSAG), the Northern Ireland Statistics Advisory Committee (SAC) and the inter-departmental Statistics Co-ordinating Group (SCG).

Censuses in Northern Ireland are governed by legislation, the primary legislation being the Census Act (Northern Ireland) 1969. The conduct of the 2021 Census in Northern Ireland also required a Census Order and Census Regulations to be passed by the Northern Ireland Assembly. In preparation for this, NISRA published a proposals document that described the planned census process from end-to-end. Alongside this a number of impact assessments were published and these focussed on the identified requirement for each question and how NISRA would ensure that census data would be kept secure.

2.2. Developing the census processes and infrastructure

The conduct of any census is complex and Census 2021 was no different. It required significant planning, development and integration of many component parts.

The content of the census questionnaire was developed continuously from the first topic consultation in 2015, in which respondents were asked about the value of the questions asked in the 2011 Census, any redundant current questions as well as ideas for additional topics. Between 2015 and 2020 (questionnaires were finalised and printed in 2020), the questions were thoroughly tested using both qualitative and quantitative techniques. The content of the questionnaire was formally legislated for by the Assembly, with the topics to be covered being listed in a Schedule to The Census Order (Northern Ireland) 2020, and a copy of the final questionnaire being included as a Schedule to The Census Regulations (Northern Ireland) 2020. In developing the Northern Ireland census questions, there was a high degree of collaborative work undertaken with the other Census Offices in the UK. As a result, the vast majority of questions were harmonised across the UK. The Northern Ireland questionnaires are available on the NISRA website.

The high-level census methodology was also developed in the period before the census. For the 2021 Census the initial contact method was post out, with the majority (80%) of addresses initially receiving an invitation letter, containing a unique access code, to enable them to complete their census via a specifically developed electronic questionnaire. The remaining addresses were issued with paper questionnaires (these too contained a code to allow online completion, if preferred).

The use of a post-out method was critically dependent on the development of an authoritative address register. The base list of properties was taken from the Land and Property Services (LPS) POINTER database, a resource that NISRA has been continually reviewing and working with LPS to quality assure and update. While POINTER formed the basis for the address records, the census address register was, where relevant, supplemented with other known addresses that officials sourced from a range of address level administrative datasets where those addresses did not directly link to POINTER.

The plan was to overpopulate the address register and allow the field force and/or the public to make corrections to the address list by informing census systems of changes or inactive addresses.

In this development phase, NISRA worked in partnership with the Office for National Statistics (ONS) to procure services to support the data collection operation. This included contracts that covered the printing of paper census questionnaires, reminder letters and support materials as well as a system to capture and code the returned paper questionnaires. A system, known as a field work management tool, was also procured to manage the non-response follow-up workloads in the field. Another significant procurement included a contact centre to assist respondents in completing their census return as well as providing a telephone capture facility to make a return over the phone.

For 2021 the internet response channel, the ‘electronic questionnaire’ was developed and delivered by the ONS - Digital Services Team (DST). This platform was jointly tested by the NISRA and ONS questionnaire content teams. Amongst other things, DST also developed the Response Management (RM) system which enabled the tracking of the status of responses and this was connected to or updated with management information from the previously mentioned supplier’s systems. Further details on the operational running of the census, as well as key facts and figures, can be found in the Census 2021 operational report published in January 2022.

NISRA Census Office worked closely with ONS throughout this development stage and were involved in all areas of the process from creating requirements to testing and running the various systems required for the operation.

As part of the development phase and operational readiness testing for Census 2021, NISRA ran a census rehearsal in October 2019. This involved an end to end test of all the systems as they were planned for 2021, just on a smaller scale, with around 19,000 households invited to take part. This ensured that all elements of the processes and infrastructure, for example, field operations and supporting public interface services, were fit for purpose. It also assured the performance of key parts of the data processing system developed by NISRA.

2.3 The wave of contact and field operation

A key part of the planning phase was to develop an enumeration strategy for the data collection operation of the 2021 Census. From this came the wave of contact and therefore the building blocks for the field operation.

2.3.1 Wave of contact

The wave of contact documented the key times and approaches that would be used to interact with the public. The plans for this were outlined in the census proposals document and, with a digital first approach along with a post out/post back model for paper, it meant that most households should have had very limited or indeed no direct interaction with census office staff. So as an initial stage of the wave of contact, an advertising campaign was developed to raise awareness of the census, across Northern Ireland.

The census is a devolved responsibility in Northern Ireland and as such NISRA ran its own advertising and media campaign with a local supplier. Ultimate responsibility for the UK-wide advertising campaign for Census 2021 lay with ONS and accordingly and accordingly NISRA worked closely with colleagues in the ONS to ensure consistency of message with the advertising and publicity campaign for the census in England and Wales. The outdoor poster campaign included photographs that were shot locally but on similar themes to those in England and Wales, while the television advertisements had similar visuals but local voice-overs.

The census operation went fully ‘live’ on 23 February 2021 when the internet online response channel opened. At the same time some 843,000 invitation letters and paper questionnaires were posted out. This is the first interaction most people will have had with the census.

Another new approach for 2021 was the use of reminder letters in the wave of contact. Non-responding addresses were identified and issued with reminder letters on 3 dates across the 6 week period following census day. As part of the second reminder a paper questionnaire, along with a new access code, was issued to households who had not completed by 19 April 2021. Census returns were accepted, no matter how late they were returned.

2.3.2 Field operation

In the run-up to the census, Census Office recruited a temporary field force of almost 1,900 staff. The field force was responsible for following up with addresses where no response had been received – (non-response follow-up). The timing of this element of the design was impacted by the ongoing COVID-19 pandemic. Non-response follow-up ran for 6 weeks after census day up until 9 May 2022. In addition, the contact centre provided support to the public in a number of ways, for example, by arranging for new access codes to be issued, for alternative questionnaires or translation booklets to be posted out to households.

Although census returns were not required until census day, returns started to be submitted from 23 February onwards. The completion of forms was monitored and, four days after census day, field staff received their workload via the fieldwork management tool on their device. Their workload consisted of a list of non-responding addresses and field staff needed to assess whether or not a return was expected from these addresses. If so, they would try to encourage the householder to comply and make a return. If a response wasn’t expected, for example, the address was derelict, field staff recorded this so that the address register was kept as up to date as possible. Over a six week follow-up period, the field staff repeatedly visited addresses from which a completed census return had yet to be received.

Further detail on the full data collection operation is contained in the Census 2021 operational report on the NISRA website.

2.4 Data capture / information assurance

Census returns were made either online or on paper. Over 80% of household returns were made online. Both the online and paper versions of the census questionnaire were designed to ease completion by respondents, with most questions based on ‘tick boxes’, and a relatively small number of questions involving ‘write-ins’. The paper version was also designed to assist in the electronic capture of the data. The information from completed paper questionnaires was captured using ‘scanning and imaging’ technology, while write-in responses were put into classifications through auto-coding software that allowed ‘fuzzy matches’. These activities were measured against pre-defined quality levels, across the operational period, to ensure capture specifications were met. Information from the online questionnaires was processed internally to an agreed specification, and merged with the coded information from the paper questionnaires to create a consolidated dataset.

Census Office is content that the service provider captured and coded the information from the questionnaires in accordance with the agreed requirements and service levels.

Census returns contain personal information provided by respondents. Census Office places a high priority on respecting the confidentiality of respondents, and the census questionnaire states that it will be kept confidential. Census Office has taken forward a number of initiatives to ensure that the information provided by respondents is kept secure.

The UK Census Offices jointly commissioned an independent review of information assurance. The review team concluded that “Overall, this assessment has concluded that both ONS and NISRA have comprehensive security programmes in place designed to reduce the risk of compromise to the delivery of the census and citizen data”. The reports of the review team are available on the NISRA website.

2.5 Analysis and quality assurance of the census returns

The coverage, quality and accuracy of the data were monitored repeatedly from initial capture of the raw data, through a number of processes, described below, to ensure that the statistical outputs from the census are of high quality and fit for purpose. The quality assurance processes are described in more detail in the census quality assurance strategy paper.

It was anticipated that some census respondents would not answer every question on the questionnaire. As indicated in the census proposals document, missing responses to questions within a questionnaire were imputed using statistical procedures. These were the same procedures as in 2011 and in line with the rest of the UK. They were based on the Canadian Census Edit and Imputation System or CANCEIS programme (a statistical programme developed by Statistics Canada for edit and imputation in survey responses).

Although every effort was made to obtain a census return from every household and individual, it is acknowledged that a small proportion of the population will not be covered on census returns. To mitigate against this, NISRA undertook two main activities. The first was to identify all non-responding households and use administrative data to identify individuals that had been missed and insert them into the dataset – something that was also done first for the 2011 Census. The second was to repeat the coverage process that was introduced in the 2001 Census to measure and adjust for under-coverage using a follow-up survey. To facilitate this NISRA conducted a Census Coverage Survey (CCS) in June 2021, based on a stratified random sample of about 16,000 (almost 2 per cent) households across Northern Ireland. This survey was conducted using trained social survey interviewers, it tried to establish the usual residents at each of the sampled addresses on census day. The information from the CCS was merged with that from the census, and statistical models developed that estimated the coverage of the census and enabled adjustments to be made for any under-enumeration.

These statistical adjustments for incomplete census returns and census under-enumeration were designed to ensure that the census represented the characteristics of the full population of Northern Ireland. The statistical estimates were subjected to quality assurance procedures to ensure that, as far as possible, they were of high quality and representative of the Northern Ireland population. The quality assurance processes are described in the census quality assurance strategy paper (published in advance of the census) and the resulting quality assurance of the population estimates is described in a separate quality assurance report published alongside this report.

2.6 Dissemination

NISRA is committed to disseminating a wide range of outputs from the 2021 Census, free at the point of delivery through the internet. An outputs prospectus is available on the NISRA website, which details the broad schedule to which NISRA intends to adhere. The prospectus details all the statistics which NISRA intends to publish; the content of these statistics has been determined through the consultation exercises detailed above. The prospectus goes on to detail how users can request, upon demand, any further specified output or analysis, noting that these may be subject to charging.

All census outputs will be subject to Statistical Disclosure Control (SDC) techniques. The purpose of these techniques is to allow robust statistics to be published for small geographic areas while ensuring that the confidentiality of individual respondents is maintained. The techniques are based upon the idea of adding a small amount of ‘noise’ to census returns, so that statistical outputs cannot be used to infer information about a person with 100% confidence, while retaining the main messages in the statistical outputs. Targeted record swapping will be the main SDC tool used in the 2021 Census. This pre-tabular technique will help to ensure that data from the 2021 Census are safe. In addition to this, cell key perturbation will be used to further enhance the safety of output tables, allowing for new levels of user flexibility, offered via the table builder system planned for Census 2021. Details on both of these SDC techniques can be found in the report statistical disclosure control methodology for 2021 Census.

Census outputs will be produced for a range of geographic units across Northern Ireland down to a small area level. Part of this dissemination will include new small areas for 2021. Digital versions of the new boundaries for 2021 output geographies will be made available in autumn 2022, prior to the first publication of outputs using the new small areas.

The dissemination of census outputs will be complemented by documents (such as this paper) that will provide background information on the census. These additional documents will cover such areas as quality assurance, definitions for first release and frequently asked questions. These supporting documents may be developed further over time as required.

3. Further information

3.1 UK liaison and harmonisation / Use of standard classifications

The three UK Census Offices have worked closely throughout the 2021 (England, Wales and Northern Ireland) and 2022 (Scotland) censuses. This close working is articulated through a statement of agreement between the National Statistician and the Registrars General for Scotland and Northern Ireland. The Conduct of the 2021 Censuses in the UK statement outlines the progress made by the three Census Offices towards UK harmonisation. Census outputs use international standard classifications.