Quality notes, issues and corrections

Details of quality notes, issues, and corrections to published Census 2021 data.

In line with Census Office’s revisions and corrections policy, thorough quality checks will take place on all census statistics which are released. Census Office may occasionally need to revise statistics and update information, including within the Flexible Table Builder.

Details of where Census Office have discovered an issue with or correction to Census 2021 data are contained on this page for easy reference.

A record of updates made to the Flexible Table Builder is also maintained.

This page also contains a number of quality notes to help users understand quality considerations of the data.

DC005: Statistical bulletin – Marital or civil partnership status and household relationships

Type: Correction

Date added: 9 May 2024

Comment

Table 3a and Table 3b in the marital or civil partnership status and household relationships statistical bulletin contain an historical comparison of marital or civil partnership status from 1971 to 2021. In aggregating census data from 2001 onwards for comparison with the historical data, people who were ‘separated but still legally married’ were included with persons who were divorced or widowed. They should have been included with the ‘married’ group because from 1971 to 1991, those persons who were ‘separated but still legally married’ were instructed to tick the ‘married’ category.

Updated documents

The relevant data tables and associated commentary in the statistical bulletin, accompanying statistical briefing and press release have been updated on the Census 2021 main statistics for Northern Ireland (phase 3) publication page.

Analysis of the differences in the statistics

Analysis has been published that shows the original statistics and the corrected statistics for each of the affected tables. The original versions of the statistical bulletin, statistical briefing, and press release are also available for reference.

We apologise for any inconvenience.

DC004: MS-A31 Country of birth (basic detail) by broad age bands

Type: Correction

Date added: 30 November 2023

Comment

Due to a production error in ‘MS-A31 Country of birth (basic detail) by broad age band’ the heading ‘Antarctica, Oceania and Other’ was mislabelled ‘Antarctica’.

Data for this table is unaffected and the error has now been corrected on the NISRA website, including bulk download files.

DC003: Household Composition Algorithm

Type: Correction

Date added: 7 September 2023

Comment

Census Office has identified some minor errors with the application of the Household Composition Algorithm in the development of the following household/family structure variables output from Census 2021:

  • dependent child
  • dependent children (household)
  • household composition - households
  • household composition - usual residents
  • adult lifestage
  • household lifestage
  • living arrangements
  • families (count in household)
  • rooms (occupancy rating) - households
  • household deprivation
  • household deprivation (housing)
  • adults in employment (household)
  • Household Reference Person (HRP)

This issue affected 61 households and 186 people and has required minor revisions to previously published main statistics tables and data within the Flexible Table Builder.  In terms of scale, no count in any affected variable classification has changed by more than 7 households or 26 people at the Northern Ireland level.

Updates to published outputs

The Flexible Table Builder was updated and the affected main statistics tables were revised at 4:30 pm on 7 September 2023.

DC002: Household deprivation and the identification of households with children

Type: Correction

Date added: 7 September 2023

Comment

Census Office has identified an error with statistics published from Census 2021 relating to household deprivation and the identification of households with children.

Household deprivation

Around 2,000 households were incorrectly classified within the household dimension measure of deprivation.

Households with children

Approximately 15,300 households (2%) were incorrectly reported to have ‘no children’ due to an issue with identifying concealed families.  This issue affects household statistics but also statistics for usual residents who are categorised by the type of household they live in.

Affected variables

  • household dependent children
  • household composition - households
  • household composition - usual residents
  • household lifestage
  • household deprivation
  • household deprivation (household)

As a result, some main statistics tables that were released in March 2023 and also some variables and ready-made tables included in the recently released Flexible Table Builder in June 2023 are affected.  No statistical bulletins or press notices are affected.

Affected outputs

Flexible Table Builder variables:

  • household variable: dependent children (household)
  • household variable: dependent children (household) – 5 categories A
  • household variable: dependent children (household) – 5 categories B
  • household variable: dependent children (household) – 2 categories
  • household variable: household composition
  • household variable: household composition – 14 categories
  • household variable: household deprivation
  • household variable: household deprivation – 2 categories
  • household variable: household deprivation (housing)
  • household variable: lifestage (household)
  • household variable: lifestage (household) – 9 categories
  • people variable: household composition
  • people variable: household composition – 15 categories

Flexible Table Builder ready-made tables:

Tables on the NISRA website:

Updates to the Flexible Table Builder

The Flexible Table Builder was updated at 4:30 pm on 7 September 2023 with the revised data.  Users who have accessed data using the affected variables are advised to rebuild/reload their tables.

Updates to the main statistics tables

The main statistics tables affected were revised, and new versions were published on the NISRA website, at 4:30 pm on 7 September 2023.

Analysis of the differences in the statistics

Analysis has been published (Excel, 176 KB) that quantifies the differences between the original statistics and the corrected statistics for each of the Flexible Table Builder ready-made tables and main statistics tables affected.

Users may also notice some very minor changes across variable classifications unaffected by this correction that relate to correction DC003 and/or the re-application of the statistical disclosure control methodology (PDF, 168 KB).

DC001: MS-H07 Industry (2-digit) of employment

Type: Correction

Date added: 19 June 2023

Comment

Table MS-H07: Industry (2-digit) of employment, published on 31 May 2023, has been updated to now include the industry category ‘98 Undifferentiated goods- and services-producing activities of private households for own use’.

Due to a production oversight this category, for which all values were zero, had been omitted from the final table layout.

This has now been corrected on the NISRA website.

DI003: Ever worked

Type: Issue

Date added: 31 May 2023

Comment

Individual question 33 on the 2021 Census questionnaire is only asked of individuals (aged 16 or over) who indicated that they were not working in the week prior to the census.  It asks ‘Have you ever done any paid work?’ with tick box options given for recording a response. 

The wording of the ‘Ever worked’ question was updated from the 2011 Census where it asked ‘Have you ever worked?’ and if yes, asked for the respondent to record the year last worked.  The response to this ‘Ever worked’ question is used in the ‘Employment History’ and ‘National Statistics Socio-economic Classification’ (NS-SeC) census outputs.

The percentage of the population aged 16 or over with a ‘No – Never worked’ response to this question increased from 10.3% (147,000) in Census 2011 to 13.4% (203,000) in Census 2021.  This appeared high when put in the context of a growing labour market.

In addition, the results of the Census Quality Survey (CQS) for this question showed a low agreement rate of 52.9% (all other questions produced agreement rates that were above 70%), indicating that there could be an issue with question understanding.  More detail on the CQS and the agreement rates for other Census 2021 questions can be found in the CQS information paper.

To examine the ‘Ever worked’ response further, comparisons were made to a similar question in Labour Force Survey (LFS). The percentages of the population aged 16 to 64 with a ‘Never had paid employment’ response to the LFS (January to March) in 2021 and 2022 were 12.4% (144,000) and (152,000) 13.0%. When the Census 2021 ‘No – Never worked’ responses are limited to those aged 16 to 64 the equivalent percentage is 12.3% (146,000).  Therefore, the Census 2021 ‘Ever worked’ results are broadly consistent with LFS results from a similar question.  More detail on the LFS and additional years for ‘Never had paid employment’ results (Excel, 26 KB) are available to download from the NISRA website.

However, where individuals with a ‘No – Never worked’ response in Census 2021 could be linked to previous census responses, it showed that around 55,000 of the 203,000 that responded ‘No – Never worked’ in Census 2021 had previously given information stating that they were either working or had previously worked.  So while the Census 2021 responses are similar to LFS, there is still some evidence to suggest an issue.

This needs further, more detailed investigation but rather than withhold the variables from Census 2021 releases, the data has been published and accompanied by this issue note.

QN005: Economic activity related statistics

Type: Quality note

Date added: 21 March 2023

Comment

The coronavirus (COVID-19) pandemic will have affected the local labour market. This will have impacted on responses to the labour market questions on the census. For example, it is possible that some people on furlough may have identified as economically inactive, instead of temporarily away from work.

Users should be aware of this when using these and other labour market statistics

QN004: Migration related statistics

Type: Quality note

Date added: 21 March 2023

Comment

It is important to note that migration related statistics from Census 2021 will have been impacted by the travel restrictions put in place as a result of the coronavirus (COVID-19) pandemic. It has not been possible to quantify the effects of such restrictions, but the statistics are still considered useful in the context of measuring migration patterns between March 2020 and March 2021.

QN003: Marital and civil partnership status, and household relationships

Type: Quality note

Date added: 21 March 2023

Comment

The census responses contained larger than expected numbers of individuals reporting to be in a civil partnership. Information from the relationship question on the census questionnaire suggested that some of these individual marital and civil partnership status question responses were erroneous.

Census Office used this relationship information to amend a small number of marital and civil partnership status question responses where an inconsistency was clearly identified. This was only possible where people lived with their partner in the same household.

Therefore, only combined estimates for the marital and civil partnership status for the groups ‘Separated (but still legally married or still legally in a civil partnership)’, ‘Divorced or formerly in a civil partnership which is now legally dissolved’, and ‘Widowed or surviving partner from a civil partnership’ are available.

Alongside this, the complexity of ‘within household’ living arrangements may mean that household relationship statistics do not match individual level marital and civil partnership status statistics in all cases.

Users should be aware of this when using these and other statistics from these two outputs.

QN002: Place of study

Type: Quality note

Date added: 21 March 2023

Comment

Census Office identified an apparent issue, where some schoolchildren or students in full-time education had missing places of study. An exercise was carried out with educational administrative data to, where possible, identify the educational institution attended.

For those records where the educational institution attended could be identified, the census record was updated with the name, address, and associated location variables to enable distance travelled to place of study to be calculated.

QN001: Sexual orientation

Type: Quality note

Date added: 21 March 2023

Comment

The question on sexual orientation was asked of people aged 16 and over and was new for Census 2021. The law underpinning the census makes it a duty for respondents to complete their questionnaire, however, the question on sexual orientation has no statutory penalty.

In general, a process called item imputation was used to model missing census responses for other questions – however, this was not applied to the sexual orientation question. Therefore, the statistics published in ‘MS-C01 Sexual orientation’ includes the number of people who did not state their sexual orientation under the label ‘Not stated’ (3.3% in total for Northern Ireland).

DI002: Accommodation type

Type: Issue

Date added: 15 December 2022

Comment

Household question 7 on the 2021 Census questionnaire asks ‘What type of accommodation is this?’ with tick box options given for recording a response.

Census Office has noted differences in classification declared by a small number of respondents for accommodation type when compared to the accommodation type recorded by Land and Property Services (LPS) for the address. This can be explained by census respondents favouring selecting ‘semi-detached’ instead of ‘terraced (including end-terrace)’ for end-terrace properties.

It is consistent with previously published Office for National Statistics (ONS) research comparing responses to the 2011 Census question ‘What type of accommodation is this?’ with data trained surveyors have reported to the Valuation Office Agency (VOA).  The ONS research identified a preference for 2011 Census respondents to define their property as ‘semi-detached’ instead of ‘end-terrace’ in approx. 3% of cases.

This issue is related to the classification selected by respondents and does not affect the total number of properties recorded.

DI001: Rounding difference between census data published in May 2022 and September 2022

Type: Issue

Date added: 22 September 2022

Comment

The main statistics demography tables that form part of the ‘Census 2021 main statistics for Northern Ireland (phase 1)’ release supersede the population and household estimates published in the 24 May 2022 release, which reported statistics solely at the Northern Ireland level.

In May 2022, NISRA published rounded Northern Ireland level population and household estimates. At that time, the dataset was still being subjected to a number of final statistical procedures; these statistical procedures have been applied to the census database, and the unrounded statistics are now available. A number of figures at the Northern Ireland level (when rounded to the nearest 100) differ from the May figures as detailed in Table 1 and 2. 

Table 1: Comparison with May 2022 rounded figures – persons
Category May 2022 rounded figure Finalised rounded figure
Usual residents 1,903,100 1,903,200
Usual residents:
30-34 years
126,000 126,100
Usual residents:
55-59 years
129,200 129,300
Usual residents:
60-64 years
113,000 113,100
Females:
30-34 years
64,800 64,900
Females:
40-44 years
62,900 63,000
Females:
50-54 years
67,000 66,900
Males 936,200 936,100
Males:
40-44 years
59,300 59,200
Males:
60-64 years
55,400 55,300
Table 2: Comparison with May 2022 rounded figures – households
Category May 2022 rounded figure Finalised rounded figure
Households 768,900 768,800