Population and social statistics transformation: 2019 progress update

In March 2014, the National Statistician recommended the 2021 Census should be predominantly online, making increased use of administrative data and surveys to both enhance the statistics from the 2021 Census and improve statistics between censuses. The government's response to this recommendation was an ambition that "censuses after 2021 will be conducted using other sources of data".

It is our mission to provide the best insights on population, migration and society using a range of new and existing data sources to meet the needs of our users. This is increasingly important in a rapidly changing policy context, where we know our users need better, more timely evidence to support decision-making at both national and local levels.

In 2021, we will continue to meet these needs largely through the census, while also increasing our use of other sources of data. As set out in our previous work on the Administrative Data Census project, a census provides granular data at the lowest levels of geography every 10 years. However, our census-based system delivers less detail throughout the interim years. Additionally, the quality of our population estimates declines as we move further away from the census year.

Therefore, we are working to put administrative data at the heart of our population, migration and social statistics systems. Enabled by data-sharing powers in the Digital Economy Act 2017 and guided by our data security principles, we are using the opportunity to make use of more data to give us a much richer understanding of how our population is changing. This report sets out progress towards achieving these aims since our previous report in June 2018.

Population and migration estimates

We have continued to work in partnership across the Government Statistical Service (GSS) to put administrative data at the core of our population and social statistics system. We have improved our administrative data-based population estimates (ABPEs), where our latest methods show promise for removing patterns of over-coverage seen in previous research.

Our migration statistics have been enhanced through applying a preliminary adjustment using administrative data from the Home Office and Department for Work and Pensions (DWP). We have reclassified these statistics as Experimental Statistics to support ongoing development and innovation.

In 2020, we will focus on delivering improved migration statistics by further refining our preliminary adjustment methods and continuing the development of our ABPEs.

Information on population characteristics

We have continued our research to understand the availability and statistical quality of information to support an administrative data-first approach to producing social statistics. Our collaborations with other government departments have allowed us to produce our first Research Outputs on fuel poverty (Department for Business, Energy and Industrial Strategy (DfBEIS)) and highest level of educational attainment (Department for Education (DfE)). We have also published administrative data-based income statistics (ABIS) as Experimental Statistics for the first time. Our future plans include updating our ABIS research and publishing a series of Research Outputs on housing characteristics from administrative data in 2020.

By making more use of administrative data, we have also improved our crime statistics as well as providing new insights into loneliness, deaths of homeless people and high streets. You can find out more about our work in our Public Policy Analysis Quarterly Review.

Survey transformation

We are continuing to develop our vision for how social surveys will support the administrative data-first approach. This consolidates our requirements for a Population Coverage Survey (PCS) (to support the production of ABPEs) with our transformed social surveys. This will allow us to:

• fill gaps in administrative data where limited information is available on topics, such as religion and sexual orientation, as shown in Figure 2
• assess and adjust for definitional differences and under- or over-coverage to produce administrative data-based population, migration and household estimates
• improve the efficiency of our surveys by introducing an online mode of collection
• reduce the burden on the public through only collecting data once and reusing these data where possible

The Digital Economy Act 2017 provides a legal gateway for the Office for National Statistics (ONS) to access data held by public authorities and commercial undertakings to support the production of official statistics and National Statistics, including the census. Access to data is essential for the transformation of our population and social statistics systems to achieve our ambition of a fully integrated administrative data-based approach in 2023. We continue to work with data suppliers across the Government Statistical Service (GSS) and in the private sector to understand the role their data can play in a transformed statistical system, but at present we do not have access to the full range of data we need to achieve this.

Our focus this year has been on working with data suppliers to develop clear, ongoing data delivery strategies, including:

• initial supplies of the main Department for Work and Pensions (DWP) benefits data; these were shared with the ONS for the first time in October 2019, with future, ongoing supplies being planned
• more detailed benefits and income data from the DWP
• Pay As You Earn (PAYE) Real Time Information (RTI) data and Self-assessment data from HM Revenue and Customs (HMRC); we received initial supplies of Self-assessment data from HMRC, with further supplies planned to meet our transformation goals
• education-related datasets from the Department for Education (DfE) and Welsh Government

We have continued to develop our understanding of data through Statistical Quality Working Groups (SQWGs), where we discuss the data and the detail of the variables with data suppliers such as NHS Digital.

We have also worked closely with the Ministry of Defence (MoD) to understand how we can use their data to improve our statistics on veterans, to support local authorities and public bodies to enact the Armed Forces Covenant.

Information on Council Tax, in addition to helping quality assure the 2021 Census, will be used to inform housing, household income and poverty and inequality statistics, to produce new statistical outputs and to enhance our current statistics.

Next year, we will continue to work with data suppliers to understand how we can use their data in our statistics, including:

• more benefits data from the DWP to measure household income better
• Tenancy Deposit Protection Scheme and record-level social rented data from the Ministry of Housing, Communities and Local Government (MHCLG) and micro-level rental data from the Valuation Office Agency (VOA) to produce new analyses on tenure
• further information on education and attainment for Wales

Data security

As we harness the power of an increasing volume and variety of data, we recognise that we are being trusted with people's data. We take our commitment to keep data secure very seriously.

The Office for National Statistics (ONS) has implemented robust security. The business processes, technical systems and services used by the ONS for the acquisition, storage and processing of administrative data conform to government and international security standards, such as the security policy framework, ISO 27001 information security management standard and the Information Security Forum best practice. Extensive use is also made of security guidance and regular advice from the National Cyber Security Centre (NCSC) and the Centre for the Protection of National Infrastructure (CPNI).

Access to administrative data is highly controlled, based upon a "need to know", the specific data, the level of sensitivity of the data, and the national security vetting required to access it. This is underpinned by user, service, system, technical infrastructure and administrative data monitoring, using advanced tools to detect malicious events and activity. Regular security reviews are performed of systems, staff and processes to ensure that the security defences implemented are operating as designed and remain effective to protect data and operations.

As part of our commitment to be open and transparent, in January 2019, we published our data and security strategies, principles, policies and management frameworks, which set out how we use, manage and secure data while providing the best standard of statistical information for the public.

Ability to link: a fundamental prerequisite for a transformed statistics system

In order to deliver fully transformed population, migration and social statistics systems based on integrated administrative data, we need to use the best available data linkage methods and have a robust understanding of the linkage quality. Until now, we have been linking data using pseudonymised identifiers¹. This restricts the methods that can be used to link data and makes it more difficult to measure accuracy in a robust and reliable way.

In our last assessment, we described a new approach to linking, to improve the quality of the links we make while preserving the privacy and security of the data. This involves the separation of personal identifiers (including names, dates of birth and addresses) from information about people's characteristics.

Working closely with our data suppliers, we have identified an approach that will enable unencrypted identifiers to be used for data linkage while maintaining the security of the data as an ONS priority. This approach does not create linked datasets of people or households with personal identifiers, and it will increase the flexibility of our methods and improve linkage quality. Improving the quality of our linkage will improve the accuracy of our outputs and can reduce bias, since the demographic characteristics of people who can be difficult to match are often those who are of statistical interest, for example, people who move to new addresses frequently.

Where only pseudonymised data are available, we are developing new, advanced methods of linkage that will minimise linkage errors and bias. However, this linkage will be less accurate than when identifiers are available. We will continue to evaluate the accuracy of our current methods and further develop and refine them.

Meeting the information needs of decision-makers and users: our progress

Population and migration statistics

Population and migration statistics underpin a wide variety of other statistics (such as unemployment rates), support a vast range of decisions and inform public debate. For example, the ability to forecast pensions, make decisions about local services and where to site new businesses.

In a rapidly changing policy context, we also know that our users are highly interested in how migration patterns are changing and what this means for society and the economy. For example, the contribution and impact that migration has on public services - such as education and health care - and on the UK labour market. This includes both the national picture but also what is happening at a more detailed regional and local level.

The current population statistics system is heavily reliant on the decennial census. While this provides granular data for small geographies every 10 years, it delivers less detail throughout the interim years. The quality of our population estimates declines as we move further away from the census year.

In our January 2019 research update, we set out how we are investigating an alternative approach to producing population stocks and flows, making best use of administrative data and using "signs of activity" to determine the resident population. For the first time, we considered how we might produce long-term international immigration flows using administrative data. We will bring these stocks and flows approaches together into a coherent hybrid model.

Figure 1: Proposed hybrid model for delivering a transformed population and migration statistics system

Source: Office for National Statistics – Population and social statistics transformation

Download this image Figure 1: Proposed hybrid model for delivering a transformed population and migration statistics system

.png (45.9 kB)

Annual population base: population stocks

Since our last assessment, we have continued to update users on how we are developing our administrative data-based population estimates ABPEs. Building on this research, in June 2019, we published our analysis using activity-based rules and records from single and linked data sources to develop our approach for producing ABPEs. Initial rules have been combined to produce the first ABPEs using this approach. These rules have largely removed patterns of over-coverage at national level seen in previous research. While this now results in higher levels of under-coverage, this is more comparable with the results seen from the census before adjustment takes place.

This provides a platform for combining with a Population Coverage Survey (PCS) to produce coverage-adjusted estimates in a similar way to the method used to successfully adjust the census.

While we have demonstrated progress towards putting administrative data at the core of our population statistics system, we need more activity data to meet our transformation milestones.

Components of change: population flows

We have set ambitious targets to put administrative data at the core of our evidence on migration and population throughout our transformation journey. A rapidly changing policy context - including the government's plans for a new immigration system once the UK exits the EU - offers us a well-timed opportunity to reflect on the best way to deliver the population and migration statistics system to best meet users' needs.

We have published a range of updates on our progress towards this goal, including our January 2019 research engagement report, which set out our approach to integrating administrative data, and feedback on these plans.

In August 2019, we made improvements to our international migration estimates by applying a preliminary adjustment based on administrative data from the Home Office and DWP.

We also published our latest Understanding different migration data sources report. This set out the conclusions of our research into the coherence of migration data sources including the International Passenger Survey (IPS) and the Annual Population Survey (APS) alongside other administrative data sources. Based on these findings, we made preliminary adjustments to the headline measures in the latest Migration Statistics Quarterly Report (MSQR) to provide our best possible assessment of migration. These adjustments only have a small impact on UK net migration, and the overall net migration trends we see over time are largely unchanged. This is because the adjustments made to the EU and non-EU figures broadly cancel each other out. To reflect our transformation journey and the development and innovation that the statistics are undergoing, we have reclassified our migration statistics as Experimental Statistics.

Our research has highlighted the challenges in using administrative data to measure population and international migration. For example, EU migration and the ability to produce analyses for sub-groups of the population, such as nationality groups and those in prisons. We will continue to make improvements to our existing international migration estimates using the currently available administrative data, but this work has highlighted the need for further administrative data to deliver a fully transformed system that meets users' needs.

Future plans

To keep users of population and migration statistics updated on our progress towards these milestones, we plan to publish regular updates over the next year. Further information is available in our Transformation of the population and migrations statistics system: overview.

Coverage surveys

As administrative data are collected for operational rather than statistical purposes, they may not include everyone who is currently resident, for example, those who have not yet registered for services, incorrect geographical location for those moving to a new house or over-coverage where people have left the country. We will need a PCS to assess these coverage errors and make adjustments to our population estimates.

Alongside our research using administrative data, we are continuing to progress our work to transform our surveys across the ONS. The strategic model we are developing brings together our requirements for a PCS with our transformation of our existing surveys.

Since our last assessment, we have undertaken testing of PCS questions that have been integrated with Labour Market Survey (LMS) questions. The response rate was just below 61%, and we need to understand what that means for bias in our estimates. To optimise the precision of our estimates using administrative data, we have developed a sampling strategy using a new stratification model based on administrative data that aims to predict over-coverage and under-coverage in the ABPEs.

Over the next year, we will analyse the collected test data and assess how the response rates achieved affect the bias in the population estimates produced. We will provide an update on our progress in 2020.

Households and families estimates

Our focus in 2019 has been on putting administrative data at the core of evidence for migration and improving our ABPEs.

We have been working to understand the differences between occupied addresses identified through administrative data and census households. We have plans to investigate further how we can best meet users' needs for household-based statistics.

We have also been testing and refining our Generalised Structure Preserving Estimator (GSPE) methods² for household size and our imputational approach for household composition. We intend to apply these methods to our next ABPEs and update our previous Research Outputs in the future.

Statistics on families are hard to produce using administrative data alone, as it is difficult to capture the complex range of family situations captured by the census and surveys. We are undertaking research to bring relationship information from the census, birth registrations and benefits data together with survey data to meet users' needs. We are also researching the best ways to understand the characteristics of people who do not live in private households. This means that in future, we should be able to provide statistics that are more representative of the population as a whole.

Population characteristics

In our last two assessments, we published a diagram outlining the quality and availability of data for producing census topics. We have updated that diagram to demonstrate our progress (Figure 2). The composite quality measure shows how close the definition of the data is to what is needed by decision-makers and users. It also shows whether there is a known error in the data.

Purple diamonds represent the topics that we have published research on so far, including income, qualifications and fuel poverty since the last assessment.

Figure 2: Quality and availability of data

Source: Office for National Statistics – Population and social statistics transformation

Download this image Figure 2: Quality and availability of data

.png (401.7 kB)

Administrative-based income statistics

This year, we have published further progress on our income output research. Our ONS Income Research Output was renamed "administrative-based income statistics" (ABIS), and these were published as experimental statistics instead of Research Outputs for the first time. The classifying of the statistics as experimental, rather than as a Research Output, demonstrates the improvements that have been made to the methodology and the income measure since the last publication in October 2017. The Experimental Statistics status still highlights to users that there are limitations to the quality and coverage of the ABIS and therefore to how these statistics can be used. We are committed to continuing to improve the quality of our ABIS.

Our latest ABIS publication saw the inclusion of new components of income such as Winter Fuel Payment and Christmas Bonus payments. We also responded to feedback from previous publications and refined our methodology, for example, making household income equivalent and using individuals as the analysis unit in a way that aligned with other published measures of income. We also:

• included a measure of net or disposable income for the first time
• provided additional income measures, such as medians and percentiles
• published a breakdown of income by different population sub-groups
• clearly highlighted the gaps in the income measure by providing a comparison with the Canberra handbook components of income
• produced a comparison of the ABIS with published survey-based household income estimates

We intend to continue our research and publish further ABIS when we have access to sources of administrative data that will allow us to include components of income currently missing from the ABIS measure, such as income for the self-employed.

Highest level of qualification

In October 2019, we published our early research demonstrating the potential of administrative data to provide information on educational qualifications, which has been collected by the census since 1961. This research is in collaboration with the Department for Education (DfE), which has provided the feasibility All Education Dataset for England (AEDE).

This feasibility research shows that the AEDE can provide high-quality information on the highest level of qualification in 2011 for individuals aged 16 to 25 years who studied in government-funded education in England. This offers insight into a large proportion of first-time entrants to the labour market and consequently an understanding of whether this group is equipped with the skills to meet market demands.

Although our early research is encouraging, we need to further demonstrate that the dataset can provide statistical information on educational attainment to the level of statistical quality that will meet users' needs. We will continue to work closely with the DfE to improve our statistics and to understand the potential for ongoing supply of these data and how we can address any differences between qualifications data obtained from census, survey and administrative sources. We plan to explore how data for Wales can be incorporated into this research and further comparisons of administrative data against the APS for more recent years.

Other population characteristics research

We have also researched what information is available on administrative sources that could provide information on Welsh language speakers, primarily focused on how these data could be used to measure the Welsh Government's Cymraeg 2050 policy for a million Welsh speakers by 2050.

We are also investigating how administrative data sources and new methods could be used to establish veteran status and produce univariate analysis on ethnic group.

As Figure 2 shows, there are some topics where there are little administrative data available, and we will need to rely on surveys to provide this information to meet users' needs, for example, religion. The Centre for Equalities and Inclusion has developed and is maintaining an audit of data available on the protected characteristics of the Equality Act 2010³. This has involved a detailed review of information recorded against each protected characteristic in each source with the latest release in December 2019. Our current focus is on improving the information that is available from the audit and considering how we can develop tools to navigate equalities data and incorporate information from administrative data.

We will also continue to monitor the availability of other characteristics in administrative data, recognising that information on these topics will be collected by the 2021 Census and surveys in the future.

Although we have made progress, we still need further research and access to additional data sources to be able to demonstrate the potential for multivariate analyses using administrative data and surveys. We are continuing to expand our work on population characteristics and will continue to research areas driven by priority user needs.

Housing characteristics

In February 2019, we published our Research Output on secured deposits held by the Tenancy Deposit Protection (TDP) schemes in England and Wales. This assessed the feasibility of using the TDP schemes' data in the estimation of the dwelling stock by tenure. Since then, we have explored a model-based approach to estimating dwelling stock by tenure and published this research in November 2019. This research does not use TDP schemes' data, but we will be investigating ways of using multiple sources of data in the estimates of dwelling stock in our future research.

In July 2019, we published Research Outputs on our joint work with the Department for Business, Energy and Industrial Strategy (DfBEIS), developing methods to improve local authority fuel poverty statistics in England.

We have worked with the DfBEIS to test alternative methods that supplement English Housing Survey data with administrative datasets compiled by the ONS. These included ABPEs, income, household composition and housing data from the Valuation Office Agency (VOA). Using a modelling approach, we have shown progress in calculating confidence intervals for local authority estimates. Providing confidence intervals as indicators of precision is important for identifying when differences in estimates of fuel poverty over time, or between local authorities, are statistically significant.

We will continue our collaboration with the DfBEIS over the next year to see whether measures of quality can also be produced for Lower-layer Super Output Area (LSOA) estimates.

We have also continued our research into understanding how VOA data could be used to enhance 2021 Census outputs and produce administrative data-based housing characteristics. Subject to data access, we intend to publish a series of Research Outputs in 2020 on:

• property type and size
• feasibility research for edit and imputation for administrative data-based statistics for number of rooms in the 2021 Census for England and Wales
• overcrowding measures (for example, occupancy rating)
• other housing characteristics (for example, central heating)
• quality information about our statistical use of VOA data

Notes for Data security, linkage and addressing users' needs

1. Pseudonymisation uses a cryptographic hash function to anonymise personal identifying information including names, dates of birth and addresses. Further information on this process is available in Beyond 2011 report M9 (PDF, 320KB).
2. We have previously published Research Outputs on ethnicity from administrative and survey data using the GSPE method.
3. The protected characteristics of the Equality Act 2010 are age, disability, gender reassignment, marriage and civil partnership, pregnancy and maternity, race or ethnicity, religion or belief, sex, and sexual orientation.

We are continuing to consult with our users via a series of events and request feedback from users in our publications. If you would like to get involved, please email SDR.Stakeholder.Engagement@ons.gov.uk.

In our research engagement report published in January 2019, we asked users for their feedback on our population and migration statistics transformation journey so far, to ensure our plans continue to meet users' needs. In addition to written responses to this report, we have also presented and discussed our plans with users at dedicated meetings, webinars, roadshows and round table events.

We summarised this user feedback in June 2019, which showed clear support for our ambition to put administrative data at the core of our population and migration statistics transformation, while acknowledging the ambitious and challenging programme of work. Users highlighted the important role that population and migration statistics have in informing a wide range of decisions, both nationally and locally (including in education, housing and health care) and welcomed the opportunities that exploring new statistical definitions and a wider range of sources could bring.

The Office for National Statistics (ONS) continues to work closely and to collaborate with National Records of Scotland (NRS) and the Northern Ireland Statistics and Research Agency (NISRA) on the use of administrative data and the 2021 Census. This includes the UK Population and Censuses Strategic Group as the main forum to discuss the harmonisation of the use of administrative data and improvements to the existing population statistics system, use of administrative data in the 2021 Census, and coordinating plans across the UK for future transformation of the population, migration and social statistics systems and improving the provision of UK statistics.

In addition, each of the four nations are represented at other groups, including the Administrative Data Research Group, which meets twice a year to share information and best practice on how to effectively integrate administrative data in the census process. At monthly Devolved Administration Joint Working Groups, the nations discuss the ONS's acquisition of data through the Digital Economy Act 2017 and how this may support the devolved nations.

We have continued to engage with other national statistical institutes on a bilateral basis, through the UN and other specially convened working groups. Many other countries recognise the opportunities in integrating data, and we continue to collaborate to help us address the challenges we face in using administrative data and transforming our statistical system.

The ONS Population and Public Policy (PPP) Annual Conference in July 2019 updated our users on our progress on our transformation journey and provided opportunities for further feedback. The 2020 conference will be an important opportunity for users to find out more about our progress and plans.

In the coming year, we intend to:

• work with data suppliers across the government to ensure we have the full suite of data necessary to deliver the system
• refine, develop and evaluate our current methods for data linkage
• improve international migration statistics for the UK
• continue research into our approach for delivering ABPEs
• research and assess the impact of including adjustments made to international migration statistics in the mid-year population estimates for England and Wales (reference year 2019)
• assess how achieved response rates could affect bias in population estimates produced using these methods, using PCS test data
• research how we can integrate relationship information from different sources to meet users' needs for statistics on families better
• research the characteristics of people who do not live in private households
• continue our research and publish further ABIS to include components of income not currently included, for example, income for the self-employed
• continue our collaborative research on fuel poverty with the DfBEIS and highest level of attainment with the DfE, exploring how data for Wales can be incorporated into this research
• publish a series of Research Outputs on housing characteristics
• monitor the availability of information on characteristics in administrative data, exploring potential for multivariate analysis, driven by users' needs
• review how we engage with our users as we move towards a transformed population and social statistics system, looking ahead to our recommendation on the future of the census in England and Wales in 2023 and moving away from the criteria developed for our previous work on the Administrative Data Census Project
• continue to hold user events on specific topics, as appropriate, and seek feedback on our Research Outputs

We will continue to update users on our progress in transforming the population and social statistics system in 2020 through our publications and user events, including a conference in July. If you would like to get involved or provide feedback, please email SDR.Stakeholder.Engagement@ons.gov.uk.