About this article
Office for National Statistics are investigating alternative methods for producing population estimates by ethnic group.
New approaches for producing population estimates by ethnic group report
Office for National Statistics (ONS) publishes detailed population estimates by ethnic group for areas in England and Wales following each census. However, there are currently no reliable population estimates by ethnic group available at the local authority level for the years between censuses. Given the user interest in regular, local authority ethnicity data, ONS has been investigating alternative methods to meet this need.
Today (25 August 2017) we have published:
- a research note, accompanied by illustrative estimates for 2016, on a method of combining data from the Census and the Annual Population Survey (APS) to produce population estimates by ethnic groups for UK countries and local authorities and higher levels in England and Wales – which begins to address user interest for more current information on ethnicity
- an approach for estimating ethnicity from survey and administrative data at the local authority level – which is an investigation into the longer-term potential of using administrative data as well as surveys
These research articles respond to a user need for more regular and robust estimates of ethnicity. In particular we’re investigating a strategic approach, using combinations of the data we currently have available and methods that can be developed and refined over time. These publications take on recommendations from users to investigate a wider range of data sources, the use of social survey sources and small area estimation techniques. Today’s publications make progress in these areas.
Formal annual estimates of the population by ethnic group were first published (for LAs in England only) by ONS in 2006. These Population Estimates by Ethnic Group (PEEGs) continued to be published as experimental statistics until 2011.
These estimates were produced using the standard cohort-component method used for the official mid-year population estimates. The cohort-component method ages forward the census population one year, then accounts for births, deaths, international migration and special populations such as the armed forces. The PEEGs extended the mid-year population estimates by using other data-sources (most notably the 2001 Census) to estimate the ethnic composition of each component of population change - births, deaths, and migration.
While the PEEGs approach had several advantages, comparison with other data sources suggested that the estimates may have drifted from the true values for some sub-national areas. For example, the 2009 PEEGs estimated that the White British group made up 60% of the population of London while the Annual Population Survey (APS) estimated the proportion as 51%. One possible cause for this difference was the lack of a reliable method for estimating the ethnic composition of internal migration (the largest component of population change for local authorities).
Since 2012, users have relied on the results of the 2011 Census as providing the best picture of the ethnic composition of the population. However, we have identified increasing user interest in updated estimates as the 2011 Census results become less current. The research articles described below reflect our progress towards producing outputs that meet this user need.
APS-based population estimates by ethnic group
We already publish estimates of the population by country of birth and nationality at local authority level. These estimates are derived directly from the Annual Population Survey (APS) and are not aligned with the mid-year population estimates. Today we have published a Population Estimates by Characteristics research paper describing a simple method of adjusting the APS results so that estimates consistent with the mid-year estimates can be produced. We also describe the use of this method to derive population estimates by ethnic group and publish a set of tables containing illustrative estimates for 2016. This method provides a quick and easy way of estimating the population by ethnic group in the decade following the 2011 Census.
An approach for estimating ethnicity from survey and administrative data
The Research outputs: An approach for estimating ethnicity from survey and administrative data research paper describes early research investigating the ability to produce estimates of ethnicity in an Administrative Data Census context. Ethnicity data is important to users, however our initial research finds that ethnicity information is not widely collected on administrative sources and the statistical quality, for the purpose of producing estimates, varies. The approach uses the Generalised Structure Preserving Estimator (GSPREE) approach to make the best use of available data, combining survey and administrative data sources to produce population estimates by ethnic group. This approach offers flexibility in terms of being able to incorporate new data sources as they become available.
As this represents an early investigation into the approach, this report focuses on the methods used and an analysis of how well they perform. Estimates are not published alongside the report at this stage.
However later in 2017, a follow-up report will use this method to produce mid-decade estimates for 2015 using an administrative data population base. This will enable further comparisons to be made between the two methods described in today’s releases.
Main differences between the approaches
The individual reports provide further details on methodological approaches, the key differences are summarised below:
Table 1: Main differences between the approaches
|Table 1: Main differences between the approaches|
|GSPREE Ethnicity Research||APS-based estimates|
|Geography||Local Authorities in England||Countries in the UK and local authorities and higher areas in England and Wales|
|Reference Date||(analysis only, estimates not yet published) 2011||2016|
|Ethnic Groups Covered||6 ethnic group categories: White, Mixed, Asian, Chinese, Black and Other. Groups chosen to ensure consistency across data input sources.||6 ethnic group categories:|
|White British , All Other White, Mixed, Asian, Black, Other.|
|(analysis only, estimates not yet published)|
|Availability of Underlying Data||2001 and 2011 Census local authority data available from NOMIS.||2011 Census available from NOMIS.|
|2011 APS data used here are available under special licence.||2016 APS data used here are available under special licence.|
|English School Census (ESC) data used here are available under special licence.||2016 Mid- year population estimates at local authority level available on ONS website.|
|Approach Summary||APS is used with the ESC and census data to produce model estimates. These estimates are then adjusted with robust row and column totals. The methods draw strength across sources, increasing the precision of the estimates. This approach can help to produce estimates for categories where sample size is small or zero.||Mid Year population estimates are split into the household and the communal establishment populations based on census proportions. Current year APS distributions are applied to the household population. Census and APS data are used together to derive the distribution of the communal establishment population. Estimates are then combined to give a final estimate.|
|Draws strength from multiple data sources.||Consistent with the standard mid-year population estimates and are based on standard classifications which make it easy to use the estimates in conjunction with other datasets.|
|Approach is able to produce estimates even with small or zero counts in survey data.||Based on existing National Statistics which have supporting quality information.|
|Possible to produce estimates for small areas.||Reflect differences in the characteristics of the household and the communal establishment populations.|
|More up-to-date than the standard source of the 2011 Census.|
|No measures of uncertainty available at this time.||Sampling error and other sources of uncertainty make the estimates less reliable than the 2011 Census results were at that time and these sources of uncertainty have a particular impact for smaller groups or when measuring change over time.|
|While information on sampling error in the APS is available, uncertainty through the new methods cannot be easily quantified.|
|Do not use the level of detailed classifications in the current population estimates by country of birth and nationality release.|
|Cannot be produced for areas or small population groups where the APS sample is too small.|
|Source: Office for National Statistics|
Download this table.xls
Feedback and next steps
We are keen to get feedback on these approaches including how they might be improved and potential uses of the data. Please email your feedback to firstname.lastname@example.org. Don't forget to include the title of the report in your response.
We plan to publish further analysis on the differences between the approaches later in 2017.