Session 1: Statistical design for the 2021 Census
Use of administrative data in 2021 Census
ONS intends to use administrative data in its design of the 2021 Census, statistical processing and production of integrated/enhanced outputs.
Response to the 2021 Census
The 75 per cent target for online returns is based on the number of household responses rather than people. The challenge of dealing with missing responses in households, particularly for online responses, was raised. ONS plans to design an online form for individual responses to assist with improving response within households, along with clear instructions and support to maximise within household coverage. In addition, the Census Coverage Survey will also provide some information on within household non-response to inform any coverage adjustment.
Improving response and coverage adjustment
Suggestions for identifying duplicate responses included asking for National Insurance Numbers (NINos) and linking multiple responses (eg 2nd addresses). ONS are not currently planning to collect NINos in the 2021 Census but are exploring how administrative data and other methods can help adjust for non-response within households. The Census Coverage Survey will also help with over coverage.
Definitions and population bases
It was noted that the definition of a household and units within them, such as families, are fundamental to the census. The definition of a household continues to evolve and change and we will be working on this to ensure that the 2021 Census will use the right definitions that relate to our society. ONS will also need to understand the definitions used in administrative data and other sources and consider potential differences in the use of these data. ONS intend to produce outputs based on definitions of usual residence and short term migrants but we also plan to release outputs for alternative populations such as daytime, work time, second residence and term-time.
2021 Census operations
Participants were interested in plans for doorstep data collection and support for completing census forms. ONS are considering how hand-held devices could be used in the field for example; to help field staff assist respondents complete their questionnaires. Of course, there are important security issues to be considered here. In addition ONS plan to provide public assistance through community support and other activities, possibly including the use of community hubs.
Session 2: Progress towards an Admin Data Census: the story so far
Use of other data sources in developing an admin based census
Much of the research to date has focussed on the use of government administrative data. It is clear that access to a greater range of administrative and other data sources is needed if the government’s aspirations for a ‘census after 2021 to be constructed using other data sources of data and providing more timely statistical information’ are to be realised (Government Response to National Statistician’s Recommendation, 2014). Data sources such as utility data and business registers were suggested as useful activity indicators.
Population characteristic estimates
Participants emphasised the need for information that measures the granularity of change (eg multivariate outputs and change over time at small area level), reinforcing the feedback from the Beyond 2011 Programme consultation. Current research plans focus on making greater use of administrative data in combination with surveys as opposed to simply relying on a survey (which was the only feasible approach at the time proposed by the Beyond 2011 programme). Development of the methods to produce these statistics is ongoing but recognised as the most significant methodological challenge for an Administrative Data Census. The ultimate goal for the Census Transformation Programme is to be able to produce the same range and level of data as the current decennial census produces but more frequently. It may well be difficult to produce the full range of topics currently collected in the census but an Administrative Data Census could offer the ability to more easily produce statistics on topics not traditionally collected in a census, such as fuel poverty or housing affordability.
Joined-up approach to producing GSS outputs
Working across Government Statistical Services (GSS) and beyond is essential to ensure a coherent statistical picture. ONS chair a Census Transformation Programme Data Suppliers Group which reviews research into using accessible data sources.
Data access legislation changes required
ONS currently has access to some but not all data required to progress. The existing legal pathway is slow and cumbersome and does not account for future changes. Proposed new legislation in the Digital Economy Bill would provide powers similar to those already held by Statistics Canada and the Central Statistical Office in the Republic of Ireland. This would include giving the National Statistician (NS) a ‘right of access to data held by public authorities and larger private undertakings… for producing statistics…’ as well as an obligation to consult the NS before changes to data collection are made. See the Annual assessment of ONS’s progress towards an Administrative Data Census post-2021 for details.Back to table of contents
1a – Administrative data research
Assessing the quality of population, household and population characteristics estimates
Delegates were keen to understand how the quality of the estimates would be measured and assessed against user requirements. Uncertainty measures around the estimates would enable an independent assessment of the quality of the Statistical Population Datasets (SPDs) and other derived estimates. ONS plans to review current quality standards, and assess how an independent quality measure (such as confidence intervals around estimates) might be produced. For population characteristics produced using a combination of administrative and survey data, a similar approach will apply.
Long term, one use for the Population Coverage Survey (PCS) will be to provide confidence intervals around the population and possibly household estimates. ONS are considering how the 2017 Census Test, the 2018 PCS Test and 2021 Census can be used to validate the administrative data outputs.
Range of household estimates
ONS are planning to produce a first set of research outputs on the number of households for 2011 and 2015 at local authority level by the end of 2016 and later will include the size and composition of households, and lower geographies. Replicating census outputs may be more challenging for topics with limited administrative data (eg tenure); ONS may need to rely more on surveys for such topics with the consequence that outputs aren’t available for very small areas. ONS will continue to explore additional data sources that may help to fill the gaps.
Joining-up ONS social surveys
This is central to the ONS Data Collection Transformation strategy – see the ‘Wider Transformation in ONS’ session, below.
1b - Outcome of the 2021 Census Topic Consultation and next steps
The session provided an update on the 2021 Census topic consultation, and ONS plans for development. The rest of the session was run as a workshop to collect information on user requirements, understanding of new topics such as volunteering and gender identity and the challenges and opportunities for a new questionnaire design. There are no slides available for this session but you can find out more about the 2021 Census Topic Consultation.
2a - Improving quality of admin data
Quality assurance of admin data toolkit
Statistics derived from administrative sources are subject to a range of potential biases, incompleteness and errors. The UK Statistics Authority has developed a quality assurance toolkit that requires users, producers and suppliers of administrative data to have a clear understanding of the quality of administrative sources, with questions such as: how do you know the data are sufficiently reliable and suitable to be used to produce official statistics? What do users need to know about their quality to use them appropriately? Crucially the toolkit creates a shared understanding across stakeholders from collection through dissemination to use.
Collaborating with data suppliers
The Census Transformation Programme is working with data suppliers across a range of sources in various ways, including secondments and loans to supplier departments and statistical quality groups, all coordinated through the ONS chaired Data Suppliers Group. The collective aim is to establish mutual benefits to both supplier and ONS, to share knowledge and build evidence for the toolkit. In the longer term, the aim is to translate this into improvements to both the methods and quality.
Data quality - a supplier’s perspective, by the HSCIC
Penalties/payments to IT providers to update administrative systems/services tend not to work. Changes that result in unusable systems must be rectified quickly. The NHS deals with many departments operating differently and these varying needs must be taken into account when making changes. There is scope for reducing manual updates in future and updating the training.
Quality of multivariate statistics
It is important not to just consider each individual administrative dataset, but also combining these datasets together to build an integrated suite of data to produce outputs to meet priority information needs.
2b – Methods for making the best use of administrative data
Coverage of administrative sources may not fit the ‘usual residence’ population definition; this must be considered when using these data for imputation purposes. The Census Coverage Survey captures the usually resident population and could help to capture or highlight these errors.
Methods to assess coverage
There is potential for two new emerging methods - trimmed DSE and graph databases - to tackle over-coverage and the quality of data linkage.
Street level outputs
ONS currently uses Unique Property Reference Number (UPRN) solely to match addresses. There is potential to produce statistics at street level using UPRN to enhance use for service provision purposes. However, producing statistics at this level will pose potential privacy/disclosure issues which could prevent production of these outputs.
3a – Opportunities for using wider data sources
Keeping the public informed of uses of their data
Being very clear about the reasons for acquiring and using the public’s information is central to public acceptability. ONS publishes Privacy Impact Assessments which outline the mitigation in place for potential issues arising from current data sharing practices. The Research and Statistics strand of the Economic and Social Research Council (ESRC) has carried out discussions with privacy groups to the get their views. Privacy groups tend to be less concerned about the use of data for purposes of research and statistics but there seems to be more concern if the data are paid for, for example, in the case of commercial data sources. ONS is working closely with the Cabinet Office and across government to understand public acceptability of the use of commercial (and other) sources of data. A Data Science Ethical Framework has been published by the Cabinet Office to guide for those looking to balance the use of new data with respect for privacy.
Changes to public services and implication for data
The likely privatisation of the Land Registry concerns those using its data. This is an example of the need to ‘future-proof’ data access agreements. The proposed legislation would give the National Statistician a 'right of access to data held by public authorities and larger private undertakings for the sole purpose of producing statistics and statistical research' and include an obligation to consult the National Statistician before changes to data collection are made in order to protect the security of data supply, and the accuracy and reliability of statistical outputs derived from these data (Annual assessment of ONS’s progress towards an Administrative Data Census post-2021 ONS May, 2016).
3b – 2021 Census collection strategy
Strategy: what's different this time; what are we testing?
The push for a predominantly online questionnaire raised questions about how ONS will engage with people to achieve responses, the use of social media and the design considerations of the user experience on different devices. ONS is already evaluating its approach to communications and running design projects to drive the questionnaire rendering on different devices. ONS has not yet ruled out the use of field staff to input questionnaire responses on the doorstep - more work is needed to evaluate this.
Maximising response for a predominantly online census
Local authority and community involvement was highlighted as a success of the 2011 Census and ONS will build on this approach for 2021. One key area to develop further will be the hard to count index which will need to consider different criteria for identifying areas/groups that need additional support and encouragement to respond. The index will need to consider in particular, the effect of moving to a predominantly online collection method.Back to table of contents
Wider transformation in ONS
Government Digital Transformation
This session provided the context for the Census Transformation Programme in the wider government digital transformation setting. The introduction led with progress to date of the Government Digital Service (GOV.UK) towards improving public access to services, collection of data and the re-use of information. Building an address register is a key example of developing this in practice. ONS presented progress to date and offered delegates the opportunity to see it in action. An address register will provide the household frame and local intelligence for the 2021 census and is important for ONS and wider government plans beyond 2021. The session ended with progress in transforming data collection across ONS particularly; moving to predominantly online collections and making best use of administrative data.
Translating addresses into households
Participants asked how ONS intends to translate addresses in the address frame into households. In 2011, the address frame was used as a proxy for households. ONS intends to use administrative data to help with this in 2021, but still needs to develop these plans and consider potential differences in the definitions used in available data and by ONS.
Non-compliance in online responses
Participants were interested in how ONS plans to tackle non-compliance in the context of online response. Response rates for online business surveys are reported to be good at present and there are plans to test this in social surveys before they go online as part of the Data Collection Transformation Programme. Dealing with the challenges of digital exclusion are important for both social surveys and of course census; the two programmes are working closely together.
- Slides from Statistics Portugal
- Slides from the Italian National Institute of Statistics
- Slides from Statistics Estonia
Like England and Wales, nearly all countries conduct a population census at least once every 10 years. The conventional approach of the enumeration of all individuals and their details using paper questionnaires and a field operation has started to evolve; some countries perform censuses using population registers and other administrative data sources, without any field operation. Over recent years a range of different census approaches have been developed by some countries, making use of register and administrative data combined with survey data and other sources. Three examples of the census journey towards 2021 were presented for Portugal, Italy and Estonia. These examples demonstrate similar challenges particularly in moving to a predominantly online data collection model, and accessing and using administrative activity data. It was clear from the presentations that having an established population register did not preclude under and over-coverage issues.
Q&A Panel Session
Delegates were asked to post questions on a whiteboard for a panel made up of:
- Ben Humbersone - ONS
- Vanessa Cuthill - Economic and Social Research Council
- Piers Elias - Independent demographer
- Rachel Leeser - Greater London Authority
- Humphrey Southall - University of Portsmouth
The answers to the questions covered at the conference, as well as those that couldn't be answered on the day have now been published in full.Back to table of contents
The 2016 Research Conference highlighted the progress made across the Census Transformation Programme. The key themes raised at the conference demonstrated the challenge of the scale and some specifics faced by ONS, both in meeting the ambitions of the 2021 Census and in assuring the use of administrative data within and beyond the next census.
We will continue to engage our stakeholders through various routes to maintain our dialogue on our main objectives, issues and challenges, involving the:
- Annual Research Outputs in the Autumn/Winter 2016-17
- Annual Assessment to monitor progress in spring 2017
- Quality feedback loops established with data suppliers
- Local Authority Administrative Data Engagement Groups
- Census Advisory Groups (CAG) and sub groups.
ONS considers this conference a considerable success and thanks everyone involved for their input. The next research conference is expected to be held in 2018.Back to table of contents