Methodological Assurance Review Panel Dynamic Population Model sub-group minutes: 30 June 2026

Published:

Attendees

Panel:

  • Ana Basiri – (MARP representative – University of Glasgow) (Chair)
  • Fred Piel – (Imperial College London)
  • Arkadiusz Wisniowski – (University of Manchester)
  • J.J Forster – (University of Warwick)

ONS presenters:

  • Salah Merad
  • Duncan Elliott
  • Aidan Metcalfe
  • Daniel Ward

ONS attendees:

  • Dominic Webber
  • Petya Kozhuharova
  • Pratibha Vellanki
  • Paula Moore (secretariat)
  • Lynn Williams (secretariat)

Apologies:

  • Joe Southam-Gisby
  • George Ploubidis

Minutes

1. Welcome and previous actions

  1. The Chair opened the meeting and asked if there were any comments on the key messages from the March 2026 meeting. As there were no comments on the key messages they were agreed.
  2. The two open actions were discussed, and it was agreed that they could now be closed as they have been uploaded to Confluence for comment.
  3. DPMSG012 – ONS to circulate viability paper for shortlisted methods to panel members for comment.
  4. DPMSG013 – ONS to circulate statistical goals to panel members for information.
  5. It was noted that, following the pause in DPM work, all previous actions had been closed, marking a reset point for the programme.
  6. The discussion highlighted the need for improved visibility of feedback, particularly comments shared on Confluence, following the panels review of DPMAP110.

2. Disaggregating national level net migration estimates using quadratic optimisation

  1. ONS outlined a method for estimating population and migration by disaggregating national migration totals using quadratic optimisation. The method reconciles population stocks and migration flows while enforcing national constraints, producing both point estimates and uncertainty.
  2. Results showed strong computational efficiency and broadly consistent estimates with existing outputs. Improvements in precision were observed for high-migration age groups, while uncertainty was generally higher than in DPM.name change Simulation results indicated strong performance where inputs are unbiased.
  3. Discussion with the panel focused on assumptions, particularly independence between data sources. Concerns were raised about completeness of results in the paper and clarity of the mathematical formulation. Further validation and testing of the method were identified as necessary.

3. Reconciliation of Local and National Demographic Accounts (RoLaNDA)

  1. ONS introduced the RoLaNDA model as a reconciliation-based framework incorporating births, deaths and administrative data, while enforcing national consistency. It was presented as a simplified alternative to DPM with flexible inputs and scalable structure.
  2. Initial testing showed coherent outputs and efficient performance. The model allows integration of multiple data sources and represents an adaptable framework for future development.
  3. Members raised difficulties around assumptions, particularly treating births and deaths as error-free inputs and potential misallocation across local authorities. Greater clarity on input data and further testing was required.
  4. The panel recommended ONS to expand RoLaNDA documentation and clarify its assumptions and inputs. The panel also recommended to include additional simulation results and undertake sensitivity analysis of the model.

Action:

PSRPSG001: ONS to share draft paper on RoLaNDA (originally demonstrated via screenshare) with panel members.

4. Disaggregation of Kannisto-Thatcher estimates for older age groups

  1. This method enhances the traditional Kannisto Thatcher approach by modelling misallocation using deaths data and introducing uncertainty through a probabilistic framework.
  2. A Dirichlet-multinomial model is used to estimate redistribution across local authorities.
  3. Discussion with the panel focused on assumptions underpinning the model, especially on whether patterns in deaths represent the living population.
  4. Concerns were raised about bias from end-of-life movement and independence assumptions. The complexity of notation was also highlighted.
  5. The panel also recommended that sensitivity analysis be undertaken on the model, including whether additional sources such as care home data could be incorporated into the model.

Action:

PSRPSG002 – Simplify notation and improve clarity on the EldEst model.

PSRPSG003 – Share Confluence comments from panel members on EldEst model with ONS presenter.

5. DPM without system models

  1. ONS presented an exploration into removing system models from the DPM to test how the approach could be simplified.
  2. Results showed instability, convergence failures and significantly increased uncertainty, with poorer agreement with benchmark estimates.
  3. System models were found to be critical for stabilising estimates and ensuring realistic uncertainty. Removing them increased uncertainty artificially rather than improving accuracy.
  4. ONS recommended that they not continue following this strand of research. The panel endorsed this approach.

Actions:

PSRPSG003 –ONS to share full paper and results of running the DPM without system models.

PSRPSG004 – ONS to respond to the comments on the papers.

6. Next steps

  1. The chair requested that the comments from the panel were to be made earlier, so that the presenters spend more time discussing their comments.

Actions:

PSRPSG005 – ONS to demonstrate how the different methods presented in this panel compare to each other, and how they fit together for the population statistics system.

7. Any other business and close

  1. It was recognised that members had been asked to review a number of technical papers for this meeting. ONS recognised as an exceptional meeting this time with the amount of information that was being sent to the panel to review.
  2. A proposal was made for each panel member to be allocated one paper when there are lots of material to review. The chair challenged this and stated that this panel should be consistent with how the Methodological Assurance Review Panel (MARP) work.
  3. The date for the next meeting will be confirmed in due course.