• 10:30 – Introduction
• 10:35 – Actions update
• 10:45 – Future of Population and Migration Statistics Update
• 11:00 – Future of Population and Migration Statistics Attributes Scorecard
• 11:40 – Longitudinal Population Dataset
• 12:00 – EAP210 Geospatial methods for Small Area Population Estimates: Proof of concept
• 12:40 – Break
• 13:30 – Statistical Disclosure Control and Differential Privacy
• 14:10 – Excess Deaths methodology
• 15:00 – Any other business
• 15:10 – Close

Panel members

• Sir Bernard Silverman (Chair)
• Ana Basiri
• Carl Emmerson
• Natalie Shlomo
• Nik Lomax

Office for National Statistics

• Sarah Henry
• Mary Gregory (Presenter)
• Owen Abbott (Presenter)
• Neil Bannister (Presenter)
• Stephan Tietz (Presenter)
• Paola Signoretta
• Merilynn Pratt
• Mike Bracher
• Iain Dove (Presenter)
• Daniel Ayoubkhani (Presenter)
• Vahe Nafilyan (Presenter)
• Tom Tarling (Secretariat)
• Susan Williams (Secretariat)
• Emily Winstone (Secretariat)

1. Introduction (Sir Bernard Silverman)

1. Sir Bernard, the chair, welcomed attendees, and noted that he needed to leave early. Carl Emmerson agreed to chair the end of the session in his stead.

2. Actions Update (Tom Tarling)

1. The Office for National Statistics (ONS) Secretariat presented the actions update.
2. Action 100: Consultation Updates
   • The ONS stated the upcoming Future of Population and Migration Statistics (FPMS) recommendation would be published in 2025.
3. Action 108 (MARP 34): Ethnicity Harmonised Standard – ONS to follow-up on recommended datasets mentioned by panel.
   • ONS is publishing a follow-up to this work via the Analysis Function website in late 2024.
4. Action 109 (MARP 35): Secretariat to share annual report draft with the chair, ahead of 30th September final copy aim, updating following this meeting.
   • The annual report for 2023/24 has now been published.
5. Action 110 (MARP 35): ONS to provide a standing update for FPMS at future MARP meetings, for the duration of the programme.
   • A standing FPMS update has been added to the agenda.
6. Action 113 (MARP 35): ONS to add guidelines to MARP papers to communicate, and standardise, need for bibliography.
   • Amendments to guidelines has been implemented.
7. Action 111 (MARP 35): ONS to circulate proposal and confirm membership of the Dynamic Population Model (DPM) sub-committee.
   • The composition of the new DPM panel will be circulated to MARP by ONS. All key items for the DPM panel will be brought to MARP. MARP members are welcome to observe the DPM subcommittee.
8. Action 112 (MARP 35): ONS is encouraged to use legal definition of a “British National” as a starting point to align definitions and statistics, and fully explain what is being measured in publications.
   • ONS gave a short update on the “British National” definition action. The chair emphasised the need for ONS statistics to align with the legal definition, both for British and Irish nationals, noting differences in measurement by different bodies.
9. ONS presented a forward agenda, with the next committee meeting in early 2025.

3. Future of Population and Migration Statistics Update (Mary Gregory)

1. ONS presented updates on the on-going Future of Population and Migration Statistics (FPMS) programme, with several publications in Spring 2025.
2. The recommendation timeline was discussed, expanding on the action update.
3. The panel emphasised clear communication would be key for the FPMS programme, and thanked ONS for their updates.

4. Attributes Scorecard (Neil Bannister, Owen Abbott)

1. ONS presented its approach to developing a scorecard to measure the current progress in developing approaches for attribute statistics that meet user needs in the context of the FPMS programme. The approaches will use a variety of methods and data.
2. The scorecard featured various dimensions of quality and example weighting values for the dimensions as illustration.
3. The panel commented that user needs weren’t always well defined, and will differ between groups, therefore there didn’t need to be competition between scoring with different weighting methods.
4. ONS clarified the scores were to indicate readiness of statistics, not to be used as a system to prioritise one over another.
5. The panel discussed quality measures, and how quality of available data might change with time and geography. It was commented the costs of data would impact availability, with highly consistent and timely data being costly to acquire.
6. ONS agreed the time and geography quality would need to be measured together and would rethink these aspects.
7. The panel asked who the audience for the scorecard were. ONS commented it was for use on internal reporting, offering more detail and insight than simpler readiness measures. The panel agreed the scoring method was an improvement over existing approaches, giving a more granular view of progress on these statistics.
8. The panel agreed the scorecard largely covered all important quality dimensions but should consider factoring in scoring for ‘accuracy and reliability’, ‘consistency’ and ‘validity’.

5. Longitudinal Population Dataset (LPD) (Stephan Tietz)

1. ONS presented the paper and invited panel comments. A key ask of the panel was whether the defined scope of the LPD was appropriate, delivering value and meeting use case demand.
2. Discussions were held around the population coverage of the proposals, and how representative of the resident population they would be.
3. It was noted the LPD will not be as timely as the Statistical Population Dataset (SPD), due to lags in death registrations and international migration data. Differences in death registrations between England and Scotland would mean delays in youth death reporting would be significant.
4. Discussion around existing issues with public health datasets were held before moving onto the complex issue of co-residency. ONS asked the panel whether it should seek to impute residents based on historical household data, and what a suitable benchmark for this would be. The panel commented the definition of “household” matters here, and that user need on what was being measured was important.
5. Following on from discussion on coverage, the panel said coverage needed to show 100% of Census records could be linked to administrative data, and that an item showing how well this could be achieved would need to be brought to MARP in the future. ONS agreed, also agreeing to update on the ‘co-residency’ question from prior discussion.
6. The panel requested ONS to consider feedback and bring this item back to a future meeting of the panel.
7. The panel commented National Statistics Institutes (NSIs) from other countries were struggling with co-residency and definitions using administrative data, and suggested collaborations may be beneficial.
8. The panel also discussed the issue of potential disclosure if the LPD becomes census-like. ONS responded that it would consider this, and the statistical disclosure control team would be involved in aspects of the LPD.

6. Statistical Disclosure Control and Differential Privacy (Iain Dove)

1. ONS presented the paper, linking to previous papers brought to the panel.
2. Discussion around the use of Laplace noise occurred, and the ONS explained its aims with the presented approach are to preserve the mean whilst avoiding negative counts. The panel determined that it would not be feasible to preserve both the mean and the median.
3. ONS clarified its approach to creating record keys and discussed tests showing the commonality of keys was within 2% of expectation. This approach made noise independent of record adjacency in data, avoiding issues with record selection.
4. Discussion moved on to protection of longitudinal differences, referencing the earlier item of the LPD, with the panel questioning the risk of exposure. ONS noted it will be examining the scale of this risk, and existing work plans include the SDC team working with the LPD more widely.
5. It was discussed whether the LPD could use a swapping method effectively. The panel considered that perturbation alone could be enough.
6. The panel highlighted a previous paper on the topic of living arrangements, EAP207, brought previously and encouraged ONS collaboration between the topics.
7. The panel were happy with the responses ONS provided. It noted that whilst the response on geographical disclosure was sensible, it was an area of concern worth keeping an eye on.
8. The panel noted the wider FPMS programme needed to be aware of requirements for statistical disclosure control, especially with regards to geographical boundary changes that happen over time. ONS agreed, and noted the SDC team would meet with internal teams and communicate this.
9. It noted that producing synthetic data was not a good approach for reducing disclosure risk while maintaining utility. The panel suggested in the long-term that looking at ways of sharing data with experts would be valuable, given the lack of synthetic data availability. ONS said it would bring a paper to MARP on this in the future.
10. The panel requested future items on this topic to factor in feedback and discussion from this meeting.

7. Excess Deaths methodology (Daniel Ayoubkhani, Vahe Nafilyan)

1. The chair, Sir Bernard Silverman, thanked the ONS for papers presented and left the meeting ahead of this item. Carl Emmerson chaired the remaining items in the meeting.
2. ONS presented a paper on methodology revision for excess deaths. The paper presented options and invited feedback.
3. During the presentation, the panel enquired whether it was understood why excess death figures in 2019 were negative. ONS commented the 2018/19 winter was mild and hence the winter deaths rates into 2019 were low.
4. Following the presentation the panel asked whether other NSIs had been contacted on the topic. ONS responded no contact except for Canada, but is aware other countries want to move to similar modelling methods. It was clarified the ONS had no requirement for consistency with Eurostat definitions.
5. The panel commented that explainability was a key requirement for excess death figures. Whilst the more complex options presented showed small gains, there was discussion whether the gains were significant enough to warrant increased complexity. It suggested remaining with the existing 5-year baseline period was beneficial for consistency in published figures, and continuing to use the projected trend as the expected number of deaths (rather than pulling forward the mean level from the end of the baseline period) would be easier to interpret.
6. Discussions covered the uncertainty around the underlying rate in the post-pandemic period, and that such statistics had come under more scrutiny in the same period.
   The panel suggested confidence intervals and measures of uncertainty may benefit data users. ONS commented that these were currently reported, however likely underestimated uncertainty, and moving to a bootstrapping approach would be better. However, the panel was not supportive of incorporating different baseline period lengths into the simulation.
7. The panel commented due to the approach used in England and Wales for reporting deaths based on date of registration date rather than occurrence, volatility around bank holidays was to be expected. Given the calendar effect shown, the panel agreed an adjustment was sensible and beneficial, and results shown by ONS supported this.
8. There was a discussion around the reasons for minor divergence from the Office for Health Improvement and Disparities (OHID)’s method for excess deaths. ONS responded that further testing of the ONS method would consider the OHID method, and that there were upcoming meetings with OHID.
9. Following the panel feedback, ONS discussed next steps with further opportunity for user feedback. OSR would also be assessing these for potentially moving them to be Accredited Official Statistics.
10. The panel requested ONS to consider feedback and bring this item back to a future meeting of the panel.

8. Any other business

1. The Secretariat informed the panel of proposed changes to the UK Statistics Authority (UKSA) website which detailed MARP papers, minutes and reports. The panel were consulted on their needs and thoughts on proposals, which would be fed back to the UKSA team.
2. Following this business, the meeting was closed.