- 09:30 – 09:35 – Introduction – Sir Bernard Silverman
- 09:35-09:45 (10 mins) – MARP Terms of Reference for 2022 – Sir Bernard Silverman/Owen Abbott
- 09:45-10:30 (45 mins) – EAP177 – Linkage of the Census 2021 and Census Coverage to Survey Demographic Index – Elizabeth Pereira
- Break opportunity (10 mins)
- 10:40-11:25 (45 mins) – EAP176 – Admin-based Floor Area Statistics – Emily Mason-Apps/Stephan Tietz
- 11:25-11:35 (10 mins) – Correspondence comments (EAP175 – Variance Estimation for Census 2021) – Sir Bernard Silverman
- Break opportunity (10 mins)
- 11:45 – 12:45 (60 mins) – Census Update – Jon Wroth-Smith
- 12:45 – 12:50 (5 mins) – Actions update – Dominick Thompson
- 12:50 -12:55 (5 mins) – AOB – Sir Bernard Silverman
- Sir Bernard Silverman (Chair)
- Prof Ana Basiri
- Dr Oliver Duke-Williams
- Dr Nik Lomax
- Prof David Martin
- Prof Natalie Shlomo
Office for National Statistics
- Owen Abbott (ONS Lead)
- Dominick Thompson (ONS Secretariat)
- Gareth Powell (ONS Attendee)
- Cal Ghee (ONS Attendee)
- Jon Wroth-Smith (ONS Attendee)
- Charlie Wroth-Smith (ONS Attendee)
- Becky Tinsley (ONS Attendee)
- Louisa Blackwell (ONS Attendee)
- Elizabeth Pereira (Presenter)
- Jack Rodgers (Presenter)
- Emma Hand (Presenter)
- Sarah Collyer (Presenter)
- Stephan Tietz (Presenter)
- Emily Mason-Apps (Presenter)
- Joe Herson (Presenter)
- Shannon Bull (Presenter)
- Andreea Butnaru (Presenter)
- The chair welcomed attendees to the session.
2. MARP Terms of Reference for 2022
- The committee reviewed the updated Terms of Reference for 2022.
3. EAP177 – Linkage of the Demographic Index to Census 2021/Census Coverage Survey
- Elizabeth Pereira presented on the linkage of the Census 2021 and Census Coverage Survey to the Demographic Index.
- The panel was concerned that ONS had not fully considered the risks of using a 2020 extract of HESA data, highlighting the modelling of students at Census 2021 was supported by admin data. The panel suggested more evidence was required to assure that it is robust to link a previous years’ HESA records with modelled Census records.
- The panel discussed the economic impact of receiving hashed datasets from the Department for Work and Pensions:
- Action 91– ONS to evidence the amount that could be saved to achieve equally accurate results with in-the-clear data from Department for Work and Pensions.
- The panel highlighted that the Longitudinal Study links Census and life events data using a source (the PDS) in the DI, for their linkage, and suggested sharing methods and quality findings from this linkage with them. ONS explained that though the linkage methodology for the Longitudinal Study is different, the linkage team is working with the Longitudinal Study.
- The panel recommended a paper by Meng (Harvard) on linkage and combining different datasets with different sample sizes and response rates, for quality assurance.
- On research question 8, the panel asked what is meant by “specific (vulnerable) populations”. ONS explained that this question seeks to encompass populations of interest to specialists in particular circumstances and requires greater clarification.
- On research question 9, the panel highlighted that OAs are designed to be as homogeneous as possible, so to be aware there may be a pre-existing correlation in characteristics between individuals found at specific addresses. The panel suggested that OA rebuilding by NISRA may provide some future insight. ONS acknowledged issues with admin data and explained that this question seeks to quantify the extent to which it mattered, from an estimation perspective, if individuals found in specific addresses (not in comparisons across an OA) are not the same if they have the same characteristics.
- The panel ensured all comments left in advance had been appropriately responded to.
4. EAP176 – Admin-based Floor Area Statistics
- Stephan Tietz presented on the approach to Admin-based Property Floor Area Statistics. Following panel comments in advance, the presentation included a log-transformed model and a revised conclusion that a stronger correlation is needed between VOA and EPC floor area data. There is a high variance in the source data (especially for smaller properties) which causes a high variance in the predicted values. At this point, the predictions are not of high enough quality to produce harmonised floor area statistics (especially for the purpose of assessing overcrowding) and further work is required to understand sources of this variance.
- The panel commented on the results of the log transformation to the final model.
- The panel recognised that there is no single source of truth for these statistics. The panel suggested exploring alternative approaches to measure overcrowding which use VOA data as a benchmark and modelled EPC data where overcrowding nears some specified boundary. However, they acknowledged there will also be a public interest in floor area statistics for purposes that are separate to overcrowding.
- The panel suggested comparing data with a sample of available floor plans. ONS clarified that the VOA effectively get their data from floor plans via planning permission, but that there remains some need to validate the data. The panel also clarified that dates within VOA data were not clear, which ONS accepted as a limitation. The panel advised against assuming the floor area across a local sub-sample of terraced housing is consistent.
- The panel discussed known quality issues associated with EPC data, citing Hardy (2019).
- The panel suggested exploring data with other agencies such as the Land Registry and Ordnance Survey. ONS clarified it had engaged with these agencies but would revisit.
The panel encouraged the publication of the findings and to explore sources of variance if possible.
5. Correspondence comments (EAP175 – Variance Estimation for Census 2021)
- The panel noted that paper EAP175 – Variance Estimation for Census 2021 has been agreed, along with amendments, by correspondence.
- ONS agreed to amend the wording in the paper submission to reflect research topics for future development.
6. Census Update
- ONS provided an update on Census operations.
7. Actions Update
- The panel agreed that the ONS will be bringing a presentation on the 2021 Census Cohort Study to MARP. Action 83 closed.
- The panel reviewed the updated Terms of Reference for 2022. Action 85 closed.
- The panel agreed that the materials for the MARP record from MARP24 were now with UKSA for publication. Actions 86 and 87 closed.
- The panel agreed EAP175 by correspondence. Action 88 closed.
- The panel agreed that the ONS will be bringing a broad study of social statistics and characteristics transformation research. Action 89 closed.
- The panel agreed that the ONS had engaged further with consultants on Iterative Proportional Fitting methods in the development of EAP174. Action 90 closed.
- The panel agreed a separate meeting took place with Jon Wroth Smith and Carolyn Watson (ONS) to discuss software assurance arising from Action 84
8. Any Other Business
There was no other business.