The new member of the panel, Carl Emmerson was introduced. Carl has joined to add economic expertise to the group.
2. Census and Transformation Update
The panel were updated on Census and population transformation work, including how the panel might be involved in the consultation and recommendation on the future of population statistics.
3. Actions Update
Action 93, to update the Terms of Reference, was noted as complete.
4. Population 24/7– EAP187
ONS gave a brief presentation outlining the key points from the paper.
The Panel asked about how events might affect the practicalities of measuring populations in the short-term. ONS said they were not looking to record very short-term changes such as events, but that the results should be representative of ‘normal’ days.
The Panel asked whether tourist and transitory populations were considered temporary and advised that ONS will need particular clarity on definitions. ONS agreed, and said they recognised the need to complete definitions.
There was some discussion of available travel information; ONS advised that there is less available information on rail/cyclist movements. The National Travel Survey (NTS) plus some other sources could make it possible to disaggregate the population beyond age/student status, perhaps to the motivations and mobility of populations. ONS agreed but would treat accuracy of the densities as a higher priority for now.
The panel suggested that mobile phone data could perhaps be best used as a supplement or quality assurance tool for other methods, given the limited access available in the UK. They also raised risks with public acceptability and reliable availability of administrative data. Further, similar work on second home data shows that the situation can be very complicated; for example, for students and shared parenthood situations. Other sources, like Wi-Fi connection tracking, were suggested – thought the issue of some methods being ‘capped’, e.g. mobile phone data with large crowds, was also raised.
ONS responded that depending on the use of the data, the limitations of administrative data were definitely being considered. Similar, ONS are thinking about user acceptability of different sources.
MARP said that for evaluation purposes ONS will need something independent of the current data, because the existing sources might be similar.
The panel also raised the need to think about status of publication (National Statistics/other), and when lower quality results might be acceptable to help statistics keep pace with societal change.
5. Fractional counting – EAP188
ONS gave a brief presentation.
The panel commented that the previously presented population estimation process appears different to the process given on this topic. ONS explained that it did describe the same broad process, but the emphasis was on different areas and so it does look rather different. ONS said that they would try to bring back a process chart that was more clearly consistent and demonstrated the coherence of the different aspects of the work.
MARP commented that they were very pleased to see the experimentation here. They asked about the scalability of the method to multiple variables, and what methods ONS might use to deal with this. The Panel also raised that because multiple methods could be used in principle, there was a need to ensure there was a clear evidence base for the method which would actually be chosen.
The panel raised a question about bias at lower and upper levels of integer vs fractional counting, which seemed unintuitive. They were also concerned about ‘combinatorial catastrophe’ in trying to constrain across too many characteristics. ONS are looking at possibilities – for example, they could make some of the characteristics independent of the process or make the person and their characteristics whole and then place them geographically fractionally. ONS could also use fewer characteristics to produce totals which a more detailed model could then constrain to.
6. Quality standards for 2023 recommendation – EAP189
ONS gave a brief presentation.
There was discussion on the interpretation of the point estimate and the measures of uncertainty. ONS explained that that census and Mid-Year Estimate (MYE) data comes from different sources, and uncertainty is calculated differently too. For MYE, bootstrapping process can give estimates outside of CIs; though this is not very well understood. The panel suggested that this could perhaps be caused by components not being sufficiently independent.
The Panel agreed with ONS summary of the work, especially that as time goes on there are likely large movements away from census. They noted however a need to be clear about linear growth interpolation assumed in the paper, and also that overall variance should not necessarily be optimised at the cost of individual areas.
7. Linked analysis of Census and CCS – EAP192
ONS showed the panel some initial results, following a previous presentation on the work being done (EAP177). The panel was very interested in the results and made several suggestions for future work.
The panel also asked how fractional counting might affect the results; noting that if a fractional counting process is used, that may prevent Boolean counting from being used.
8. AoB
The panel noted the other papers submitted and gave some feedback detailed points.
Try to have in-person meeting next time. Early April for next meeting?
