Panel Members

• Sir Bernard Silverman (Chair)
• Dr Oliver Duke-Williams
• Dr Nik Lomax
• Prof David Martin
• Prof Natalie Shlomo

Office for National Statistics

• Julie Stanborough (ONS Lead)
• Owen Abbott (ONS Deputy Lead & Presenter)
• Cal Ghee (ONS Attendee)
• Jon Wroth-Smith (ONS Attendee)
• Helena Rosiecka (Presenter)
• Ed Dunn (Presenter)
• Leone Wardman (Presenter)
• Steve Smallwood (Presenter)
• Gareth Powell (Secretariat)
• Rob Turner (Secretariat)

1. A71 – Owen Abbott – Plan out the timetable and work areas for assurance required from panel on the 2023 recommendation, including discussion of required panel expertise.
2. A72 – Owen Abbott – ONS to bring item on overall under-enumeration strategy for Census 2021

1 – Chair’s Introduction

Sir Bernard welcomed the panel and ONS attendees. The panel discussed organisation and membership going forwards and recommended looking into what expertise may be needed for providing assurance on the work leading to the 2023 National Statisticians Recommendation on the future Population Statistics System.

A71 – Plan out the timetable and work areas for assurance required from panel on the 2023 recommendation, including discussion of required panel expertise.

A72 – ONS to bring item on overall under-enumeration strategy for Census 2021

2 – Correspondence received concerning sex question on the Census

The panel acknowledged two letters they received concerning the 2021 Census sex and gender identity questions. The panel will respond to these promptly outside of the MARP meeting.

3 – Review of Actions

No actions were closed this month.

4 – COVID-19 Update

ONS updated the panel on current work undertaken to mitigate and prepare for effects of COVID-19 on Census. This included work on communications campaign, onboarding of staff, war-gaming of contingencies, logistics of planned events, and the effect of the pandemic on under-enumeration of challenging population groups.

No actions given

5 – EAP143 – Address Centric Administrative Combined Intelligence Dataset (ACID)

Background

The Address Centric Administrative Combined Intelligence Dataset (ACID) is a proof of concept project that has been set up to enable work on census low response contingencies based on a number of possible scenarios.

This project aims to investigate how to compile intelligence from various administrative data sources linked at address level to explore the relationship between administrative and census records and help make an adjustment for non-responding households. It also aims to identify non-responding addresses on the census address frame and addresses where there is good evidence of current occupancy, to understand their likely household structure.

The exploratory proof of concept presented will be used to determine whether ACID is a viable product to help respond to contingency scenarios.

Discussion and Suggestions

• The panel discussed the implications of using the school census for the dataset. It was noted that private school pupils are not included in school census. ONS will look into spatial distribution of private and boarding school pupils.
• The panel suggested liaising with Royal Mail, who hold unique data on the individuals to whom mail is delivered, as well as being the initial provider of address data used by Ordnance Survey. ONS noted that talks with Royal Mail are ongoing.
• The panel discussed the issue of data availability and encouraged the ONS acquisition team to continue to pursue stakeholders for required administrative data.
• The panel complimented the modelling approaches used in the paper.
• ONS informed the panel that a clerical review for 2020 data is underway and will feed back on the impact of this once complete if relevant.

No actions given

6 – EAP144 – Framework for assuring the Quality of Administrative Data for Use in the 2021 Census of England and Wales

Background

EAP144 presents a Quality Assurance (QA) Framework for assessing the quality of administrative data to be used in the 2021 Census of England and Wales, based on a qualitative scoring system. Application of this framework should allow quality issues to be detected and specific metrics applied systematically, so that acceptable thresholds can begin to be developed.

The framework aims to provide confidence that administrative data are of sufficient quality to be integrated into the Census design as intended. The framework will also be useful for the QA of    administrative sources for use in statistical production beyond the Census.

Discussion and Suggestions

• When discussing linkability of data, panel suggested need to clarify distinction between variables given in-the-clear or hashed data. ONS agreed to review this section.
• The panel complimented the comprehensive look at assurance methods / indicators and suggested looking at the way these can be combined. It was suggested that weighted combination of output scores could be used to evaluate overall data quality.
• The panel expressed concern on the continued use of hashed data, suggesting possible alternatives such as matching by data providers.
• The panel noted that EAP144 has implications not just for Census 2021, but also for the broader census transformation project.
• The panel discussed the gap between the assessment of admin data and its use for the census, noting that the quality of data is dependent on how it will be utilised. ONS are working to clarify the link between this framework and census quality assurance.
• The panel suggested the use of a case study, or illustrative examples as an Annex to the framework might help guide the user.

7 – EAP145 – Performance of 2021 Coverage Adjustment Method with Lower Response and Contingencies

Background

The census coverage adjustment amends the unit level census database to make it consistent with the population estimates derived from the coverage estimation process. This paper covers the use of the Combinatorial Optimisation (CO) method for adjustment with lower levels of person and household response, and contingencies that may be required if the level of response impacts the performance of CO.

The purpose of the analysis is to assess the performance of CO when the person and household response rates were lower than expected. Simulations were set up where different amounts of persons and households were removed from the census database for various local authorities. The performance of CO in each scenario was compared to the performance with the original 2011 data.

Discussion and Suggestions

• The panel discussed removal of households to simulate low coverage and asked about the removal process, which removed households where the probability of response is low. ONS noted that this was a simplified simulation, and the next iteration will be selecting records into a simulated census rather than removal.
• The widening of the donor pool for CO using similar local authorities was discussed, as there were concerns about the high levels of donor re-use. Panel suggested that it would be useful to determine in advance which local authorities are similar and plan to implement that wider donor pool.
• The panel expressed concern about the distortion in household size, especially if there is an increase in the level of missing data. ONS responded that this will only be a concern if the ratio of people missing in counted households and entirely missed households is significantly different from the 2011 Census. Panel and ONS agreed that expanding donor pool into multiple local authorities would be a sensible mitigation strategy.
• Panel discussed potential for missing persons in households, in particular for the sixth person in a household who in 2011 were included in the primary census form but are now responding on a continuation form.

No actions Given

8 – Impact of revised Scotland date on UK Census Outputs

Discussion

ONS presented on the impact of the revised Scottish census date upon UK population outputs. No significant impact is expected on UK census harmonisation or standard tables for 2021 Census outputs for England and Wales. Potential issues surrounding the different reference dates of each census were discussed, specifically for production of special migration and special workforce statistics for the UK, as produced for the 2011 Censuses.

Discussion and Suggestions

• Panel suggested that affected UK outputs could be experimental rather than National statistics, particularly if integrating with administrative data. This could be a solution to issues of methodological assurance of these outputs while meeting user demands.
• Panel raised possibility of changing guidance for students away during term time to gather more data about students living England who attend university in Scotland. ONS agreed to investigate but thought this would be difficult at this stage.

No Actions Given

9 – Any other business

No other business was discussed