• 10:30 – Welcome and introductions – Sir Bernard Silverman
  • 10:35 – Actions Update – ONS
  • 10:40 – Future of Population and Migration Statistics (FPMS) Consultation – ONS
  • 11:10 – Estimation of Travel to Work Matrices – Li Chen
  • 11:55 – Break
  • 12:15 – Population stock estimates case study for 2021 and future directions – Eleanor Law
  • 13:00 – Any other business and end of meeting

Panel Members

  • Professor Sir Bernard Silverman (Chair)
  • Professor Ana Basiri
  • Dr Oliver Duke-Williams
  • Carl Emmerson
  • Professor Nik Lomax
  • Professor Natalie Shlomo

Office for National Statistics

  • Owen Abbott
  • Jennet Woolford
  • Sarah Henry
  • Charlie Wroth-Smith
  • Katie O’Farrell
  • Eleanor Law (ONS presenter)
  • Amy Large
  • Ceejay Hammond
  • Mark Linton
  • Isaac Shipsey
  • Li Chen (ONS presenter)
  • Kaveh Jahanshahi
  • Alice White
  • Charlotte Hassell
  • Gareth Powell (ONS Secretariat)
  • Tom Tarling (ONS Secretariat)
  • Cristina Spoiala (ONS Secretariat)

1. Welcome and introductions

Sir Bernard welcomed the group and confirmed those in attendance.

2. Actions Update

Action 96, ONS was actioned to bring back a revised Travel and Tourism proposal with the suggested investigations and further developments to MARP. Action was fulfilled by Gareth James who provided a paper for additional feedback via correspondence in August 2023. Action closed with a note that the project will come back to MARP with further methodological work in 2024.

Action 95, work is ongoing.

3. Future of Population and Migration Statistics Consultation (FPMS)

The MARP secretariat gave a brief overview of the draft report and the panel discussed what they would like to include. The ONS secretariat will redraft and share with panel members following the meeting, so that the panel can suggest further changes and provide sign off.

4. Estimation of Travel to Work Matrices – EAP203

ONS presented the work.

The panel asked about why distance was the key variable measured instead, noting that travel is about time and other factors, e.g. mode, socioeconomic background and work type. ONS responded that distance is a good proxy as a starting point but that there is a plan to replace distance with time and cost. ONS is also to consider variations in Value of Time by socioeconomic characteristics, modes and distance.

The panel also commented that the double constraint model assumes one start and one destination, yet the picture is more complex in a spatial interaction model. It was suggested that the model should be adjusted at the destination, and to take account of multiple destinations. Panel remarked that the sampling strategy was good. ONS responded that the current model follows the Census Travel to Work definition and focuses on population travelling to a fixed workplace.

A conversation on whether the display of production map occurs before or after a double constraint model was had. ONS explained that it was after a doubly constraint model. When the double constraint model finishes because convergence is achieved, predicted attraction should be very close to the one before inputting the double constraint model. However, the level of this depends on the goodness of fit for calibrating cost function which is planned to be further improved through segmentations and incorporating cost and time.

A concern was raised by the panel on the validation being based on the 2011 Census. If possible, this may be a good opportunity to see if the double constraint model arrives at observed data, which is a good indication for the validity of the approach. ONS agreed to this and responded that this requires improved cost functions and segmentation, but this is a helpful comment and something we should focus on before validating 2021 estimations. ONS mentioned that they would consider mobile phone data and Census rehearsal data as potential sources for validation.

A further concern was raised by the panel about the work and future identification of trip ends (origins and destinations) by occupation or by any other demographic. The difficulty in estimating occupation by other variables may imply that other low-level outputs are also difficult. ONS agreed and discussed the need for close collaboration with the ONS population statistics and generating bespoke datasets linking demographic and business indexes, with the aim being using both indexes.

The potential for other data sources was discussed, as well as some potential issues and opportunities.

5. Population stock estimates case study for 2021 and future directions – EAP202

ONS presented the work. Panel remarked on issues about the Statistical Population Datasets (SPD) coverage.

The panel pointed that the coverage patterns in the admin data should be similar across sexes at those young ages, randomly allocated across the country and not different, as seems to be the case. Author remarked that they had no time to explore this before the meeting and they will look further into this.

The panel also said that it would be important to estimate bias as well as variance, though it was noted that this would be difficult as we move further away from the last census.

The panel stressed the importance of quality standards, to help judge whether estimates are ‘good enough’. ONS agreed; quality standards were set out in a previous paper and will be published. This will then imply the size of the coverage survey that would be required. An SPD approach may need to be compared to the quality of estimates made at a point in time following census.

The panel queried the use of trimming, and ONS responded that they plan to do simulations of surveys. The simulations should give a feel for how the methods behave under different data scenarios.

The panel asked about the next steps. ONS said that they will look at modelling overcoverage with preliminary results later this year. Another strand of work is to support the business case for future of population and migration statistics (FPMS), informing what a coverage survey needs to look like, with testing planned for beginning of 2024. Finally, ONS are also looking at implanting the PECADO approach with inhouse data, breaking down the SPD in parts.

6. Any other business

Next meeting is scheduled for 8 February 2023.

With no other business, the panel closed.


The ONS ‘Estimation of Travel to Work Matrices’ to bring back another paper to MARP once further research is complete.

The ONS ‘Population stock estimates case study for 2021 and future directions’ team to bring back another paper to MARP once further research is complete.