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)
• Cal Ghee (ONS Attendee)
• Jon Wroth-Smith (ONS Attendee & Presenter)
• Alison Whitworth (Presenter)
• Rob North (Presenter)
• Iva Spakulova (Presenter)
• Daniel Ward (Presenter)
• Aidan Metcalfe (Presenter)
• Jonathan Rees (Presenter)
• Greg Payne (Presenter)
• Almas Mumtaz (Presenter)
• Rosalind Archer (Presenter)
• Orlaith Fraser (Presenter)

1. Jon Wroth-Smith – Action 79 – ONS to bring paper detailing outcomes and quality assurance processes for each resolution path following address checks.

1 – Introduction

1. Sir Bernard welcomed panel and ONS attendees.

2 – Actions Update

1. The panel agreed to close actions A53, A54, A67, A74, A76 and A77
2. The panel requested progress updates on outstanding actions at MARP16.

3 – COVID-19 Update

1. ONS updated the panel on current COVID-19 workstreams, in particular the implications of decision for field officers to not cross threshold of property. Panel suggested having discussions with large providers of sheltered accommodation to create a policy for field officers.

4 – EAP150 – Fractional Counting

1. ONS presented an introduction and initial proof of concept to the fractional counting approach for producing administrative based, multivariate population outputs
2. The panel discussed the decision to use oversampling to decrease the discrepancy in class sizes and suggested that this might be the cause of underestimation by the model. ONS clarified that they had also investigated an underestimation approach and had acquired comparable results, and that the unequal distribution of underestimation across each age range implies that this is not the case. The panel suggested that this could be due to the initially unequal age distribution.
3. The choice of model algorithm was raised by the panel, with the suggestion that other options would be worth investigating, such as gradient boosting.
4. Concerns were raised by the panel at the use of the term “address-centric” when describing data at postcode level.  ONS will clarify in future work, as linkage is performed at address level but fractional counting itself only at postcode level.
5. Although the current approach is an initial proof of concept, panel provided suggestions for possible improvements at a later date. These included suggestions on incorporating multivariate characteristics, a longitudinal approach to data, survival modelling and the imputation of missing data.

5 – EAP151 – Population estimation without a Census (update)

1. ONS updated progress on population since this was presented at MARP05 (EAP129 and EAP130). The aim of this work is an admin-data based estimation methodology to contribute to 2023 decision and Admin based population estimates.
2. The panel highlighted the potential for a shrinkage parameter to account for over-coverage, as used in Census Dual System Estimation. ONS agreed that this and other methods used in Census over-coverage work could be incorporated.

6 – EAP152 – Address checking

1. ONS updated panel on quality checks of the address frame. Due to COVID-19 restrictions, the previously undertaken field officer address checks were not possible and EAP152 details the steps that have instead been taken.
2. The panel discussed issues from clerical checks for houses of multiple occupancy, and the difficulties in distinguishing these from multiple distinct households.
3. ONS informed panel that utility data is being used in address checks, in particular looking at areas where there is a large number of first-time customers to discover areas with substantial amounts of new housing.
4. The panel suggested potentially feeding back information to GeoPlace to incorporate fixes into future address products, but ONS stated this is not possible as Census data cannot be used for this purpose.
5. The panel requested more information on outcomes for an address following each outcome of checks outlined in the paper. ONS agreed to do this, as well as detailing the quality assurance processes for each resolution path.
6. The panel suggested that tables 1 and 2 in the paper need amending to clarify what percentages represent.
7. ONS said that overall the new address checking processes have been an improvement on the previous method due to an increased ability to repeat checks for difficult addresses. The panel suggested assessment of address quality could also be determined using any quality reports that GeoPlace may have created, and suggested that ONS contact GeoPlace to see if these exist.

Action 79 – ONS to bring paper detailing outcomes and quality assurance processes for each resolution path following address checks.

7 – Presentation – Geodemographic analysis

1. ONS presented work on a tool to examine geodemographics during Census live operations to determine where additional resource may be required.

8 – Any other business

1. No other business