Quality Assurance report – Unrounded population estimates and ethnic group, national identity, language and religion topic data

This report accompanies the publication of Scotland’s Census 2022 unrounded population estimates and ethnic group, national identity, language and religion topic data. It builds on the quality assurance report published alongside rounded population estimates in September 2023. 

Information on response and imputation rates for this topic have also been published on the Scotland’s Census website. 

Quality information for Ethnic group, national identity, language, and religion variables

Topic expert quality assurance panel 

NRS invited analytical colleagues from Scottish Government to participate in a quality assurance panel on the ethnic group, national identity, language, and religion data. Panel sessions assessed the coherence of 2022 Census statistics with previously published census estimates for Scotland and England and Wales as well as with other available data on that topic. The panel concluded that they were content for the data to be published. 

Gaelic and Scots language skills 

The presentation of questions both online and on paper was improved for the 2022 census following phases of question testing. These differences should be taken into account when comparing results between 2011 and 2022 censuses. The full 2022 and 2011 household questionnaires are available on the Scotland’s Census website. In addition to this, more detailed guidance was provided in 2022 than 2011, with the guidance for Scots skills being the most accessed question guidance. The impact of this on the results was limited as the guidance was available on a separate webpage and was accessed by a relatively small proportion of those responding. 

English language skills 

The 2022 census English language skills (understand/speak/read/write) question included a proficiency matrix (very well/well/not very well/not at all), rather than skill/no skill as in Census 2011. In the published data we consider individuals who ticked very well or well to have the skill. We have created an additional category of ‘Limited English skills’ if an individual has ticked ‘not very well’ in at least one skill and ‘not at all’ in all the other skills. A final category ‘no skills’ refers to individuals that ticked ‘not at all’ for all four skills. When comparing to the previous census please note the change in question when drawing conclusions.  

Main language  

The 2011 census asked about ‘languages other than English used at home’ whereas the 2022 census asked about main language. We have concluded that these questions did not gather comparable data, with many more individuals in 2011 indicating an additional or alternative language to English used at home. The change to this question was based on our assessment of data collected in 2011 and question testing ahead of 2022.  

Ethnic Group 

Whilst the main ethnic group categories have not changed, some of the detailed response options and write-in prompts in 2022 were changed based on stakeholder engagement and subsequent question testing. More information on this is available in our developing census questions webpage. 

In 2022, only one tick box was available for African or Caribbean/Black ethnicities. In 2011 there were two (for African) and three (for Caribbean/Black).  This has led to large increases in ‘Other African’ and ‘Other Caribbean/Black’ between 2011 and 2022. We advise that users compare the sum of all African or Caribbean/Black categories when comparing 2011 and 2022 data.   

There were also some changes made to the tick-boxes under the White category since the 2011 census. This included the addition of a tick-box for showman/showwoman and one for Roma. These changes mean that the data for Gypsy/Traveller are not fully comparable with data from the 2011 census. This is because many people in the showman/showwoman ethnic group and some in the Roma category in 2022 may have ticked Gypsy/Traveller in 2011. 

Following feedback from stakeholders, the 2022 census questionnaire included a tick box for the ethnicity ‘White: Roma’. During quality assurance, we found a higher proportion of individuals born in Italy identifying as Roma than expected. Around 36.5% of the Roma population reported their county of birth as Italy, compared with only 19.1% born in Romania, the second most common country of birth. In total, this amounts to 1,200 people born in Italy out of the total 3,200 Roma population. Also, the distribution of Roma by local authority was different to our expectations of the Roma population being highly concentrated in Glasgow City. Only 28.1% of the Roma population are in Glasgow City, fewer than the 28.3% in City of Edinburgh. We believe the box may have been ticked in error by individuals born in Rome, Italy but cannot conclusively determine what proportion of individuals may have inadvertently ticked the wrong ethnicity, as some are likely to be people who would be considered part of the Roma ethnic group. We have decided to publish this data despite this quality issue, but we recommend that users consider this when interpreting results for the Roma population.  

Country of birth 

Users should be mindful of changes in EU members and accession states between 2011 and 2022. This will affect the number of countries which make up certain categories when comparing the results between censuses.  

Religion 

Religion is a voluntary question and 6.2% of the population did not provide a response. Please be aware that when we state percentages these are out of the whole population, not just those that provided a response. Our approach to imputation is also different for voluntary questions. Not stating a religion is considered to be a valid response, so we do not impute a religion for those who responded to the census but did not answer the religion question. However, we do impute religion for those who did not respond at all to the census. 'Not stated’ is one of the values that can be imputed for religion. More information on our edit and imputation method is available on the Scotland’s Census website. 

Quality information for population and household estimates

Residence type 

Residence type provides information on whether people live in a household or a communal establishment (CE). The response rate for people living in households was 90.5%, while for people living in CEs it was 42.7%. Administrative data (data collected about services such as how many people live in prisons or in care homes) has been used as the basis for coverage adjusting these estimates to ensure that the statistics represent the total CE population by age and sex. The quality of administrative data for CEs gave us confidence in making larger adjustments than were required for the household population. More information on this is available in the Methodology Enhancements to Secure High Quality Census Outputs and Population Estimates paper on our website. 

Comparison with small area population estimates (SAPE) for 2011 datazones 

The 2022 census population estimates for 2011 datazones can be compared with the SAPE, which have been rolled forward from the 2011 Census. These population estimates are published every year, adding births, subtracting deaths and applying net migration to the previous year’s population. Reasons for potential differences between census and population estimates for Scotland and council areas are available in a reconciliation report published on the NRS website. Figure 1 is a scatterplot, showing the 2022 census population estimate and SAPE estimate for each 2011 datazone, shown as a dot on the chart. If a dot lies on the red line, then the 2022 census population estimate and SAPE estimate are the same. 

Overall, the 2022 census estimate for 78% of datazones is within 10% of the SAPE. Some of the areas with larger percentage differences include datazones that contain large student populations, defence establishments and areas that have seen recent large-scale housebuilding. 

Comparison with household estimates for 2011 datazones 

The 2022 census household estimates for 2011 datazones can be compared with the household estimates, which are based on Council Tax data and information from the 2011 census. These household estimates are published every year, using updated administrative data. Overall, the 2022 census estimates are within 10% of household estimates for 85% of 2011 datazones. Figure 2 is a scatterplot, showing the 2022 census population estimate and household estimate for each 2011 datazone, shown as a dot on the chart. If a dot lies on the red line, then the 2022 census population estimate and census-day household estimate are the same. 

The datazones where the census-day household estimates are much higher than the census estimates tend to be areas with large student residences. Sometimes the flats within these are counted separately for council tax purposes, data which feeds into the household estimates. However, in 2022 census, these have been enumerated as a single communal establishment rather than separate households. Areas with larger percentage differences where 2022 census is higher tend to be areas that have seen large-scale recent housebuilding. 

Small changes to census population and household statistics published in September 2023

Population 

Small changes due to further census data processing since September 2023 have meant that the Scotland level population figure has increased by 3,242 people (0.06%) from the initial estimate published in September 2023, increasing Scotland’s population from 5,436,600 to 5,439,842. 

The changes were almost exclusively related to the communal establishment (CE) population. Further processing and quality assurance added in a small number of CEs and a small number were removed as further investigation revealed they were duplicates or included people who were not usual residents. These small inconsistencies were not found during the period of pre-release quality assurance involving local authorities in Summer 2023.  

By sex, the number of females has increased by 0.09% from the estimate published in September 2023, while the estimate for males has increased by 0.03%. The reason for there being more females is because most of the CEs not included were student residences with higher numbers of female than male residents. 

The increase in the population estimate from September 2023 also varies by age band. The change for most age bands is less than 0.5%. The biggest increase within an age band is 1.3% for those aged 20-24, whereas the biggest decrease is 1.0% for age 90 and over. Figure 3 shows the differences from the September 2023 estimate by 5 year age band. 

The difference from the estimates published in September 2023 is less than 0.1% for 24 of the 32 local authorities. Some of these differences are due to rounding, but they also include changes to the communal establishment population. The biggest increases from the September 2023 estimates are in Dundee City (+0.41%), Fife (+0.38%), Shetland Islands (+0.37%), City of Edinburgh (+0.36%) and Perth and Kinross (+0.10%). The biggest decreases are in Na h-Eileanan Siar (-0.23%), Orkney Islands (-0.14%) and Moray (-0.12%). Figure 4 shows the difference for all local authorities. 

Households 

The number of households remains unchanged at 2,509,269. The figure published in September 2023 was 2,509,300, which was rounded to the nearest 100.