Chair: Sir Bernard Silverman
13:00 – 13:15
|Introductions, actions review||Pam Everett|
13:15 – 14:15
|Resolve Multiple Response Methodology|
|14:15 – 14:30|
14:30 – 15:30
|Household Response Profiles|
15:30 – 15:45
|Any other Business||Pam Everett|
External Panel Members
Sir Bernard Silverman (Chair)
Prof Natalie Shlomo (External Panel Member)
Prof David Martin (External Panel Member)
Dr Nik Lomax (External Panel Member)
Dr Oliver Duke-Williams (External Panel Member)
Office for National Statistics
Pam Everett (Vice-Chair)
Sir Ian Diamond (ONS Panel Member, National Statistician)
Cal Ghee (ONS Panel Member)
Owen Abbott (ONS Panel Member)
Sarah Henry (ONS Panel Member)
Jon Wroth-Smith (ONS Panel Member)
Victor Meirinhos (Presenter)
Bethany Fitzgibbon (Presenter)
Lawrence Dyer (Presenter)
Steve Rogers (Presenter)
Orlaith Fraser (Presenter)
Pratibha Vellanki (Presenter)
Christopher Lydiat (Secretariat)
Gareth Powell (Secretariat)
|[7,3]||A64 – The panel requested an update with more detail, via correspondence.||Victor Meirinhos|
1.1 – The Terms of Reference alterations were agreed upon via correspondence, closing A63.
Resolved Multiple Response (RNR) is a mechanism to resolve duplicate census results, which may involve removal of duplicate results. The methodology is based on the 2011 Census RMR methods, with changes influenced by alterations in methodological aims, and advances in both statistical techniques and computing power.
Discussion & Suggestions
2.1 – The intention to classify missing responses as prefer not to say was challenged by the panel, as the current accommodation of legislation produces issues by classifying missing data with ‘prefer not to say’ in reported results.
2.2 – The panel highlighted the need to ensure all references are accessible. ONS agreed on this and committed to publishing referenced papers when necessary.
2.3 – The reliance on the UPRN for the definition of a household was discussed, as relying solely on UPRN could lead to issues such as the merger of households in homes of multiple occupancy (HMO) among others. ONS explained that individuals within a household are assessed before establishing if entire households should be merged. In addition, when considering merging households business rules are used that include factors other than the UPRN. This prevents the merging of households in HMO cases.
2.4 – The panel asked about the intention to make alterations to the rules during live processing. ONS responded that whilst there was flexibility in plans to allow for change, the method is considered the optimal approach for 2011 data, with a substantial review process in place to moderate possible change requests.
2.5 – How residuals are resolved was discussed, with the assignment of individuals to households described by ONS as based on data where available, but randomly should no data on which to establish which household individuals belong to exist. ONS also stated that these represent a tiny fraction of results and would only affect areas below Local Authority level.
2.6 – A question into the QA process of the algorithms was asked by the panel, with ONS responding that unit testing on all rules was in progress.
2.7 – Finally, a question on risks was asked, with the panel investigating the consequences of business rules not performing as expected. ONS responded that the number of rules was small, and the ONS was aware of the potential issues using business rules creates.
No actions given
Household Response Profiles is a best estimate of the volumes of responses during the census period. These estimates are used to influence operational decisions during the 2021 Census. The models were described by ONS, along with contingencies in place to account for potential issues.
Discussion & Suggestions
3.1 – A question into use of international approaches was asked, with ONS explaining regular discussions occur with other national statistics institutions, with expertise shared. A limitation on this sharing was noted, as differences in data available influence what approaches are most appropriate.
3.2 – The panel believed the work to produce the models and contingencies was a suitable approach but requested further detail on the methods.
A64 – The panel requested an update with more detail, via correspondence.
4.1 – An overview of the COVID Infection Study (CIS) was presented to the panel, and its interaction with the census highlighted. The panel discussed the first stage of the study as the primary source of interaction with the census, as it involves a sample of up to 50,000 households per month. Possible conflicts between the CIS and the CCS was discussed, with ONS aiming to ensure possible conflicts are avoided or managed.