4. Barriers to participation - Public trust

The Inclusive Data Taskforce report emphasised the importance of building trust: “An important message we heard from participants in our consultations was the need for a trustworthy system, supporting all groups across society to understand both the benefits and the risks of data sharing.” Over time the ONS has built an enviable reputation for trust and this is reflected in particular in the very high levels of cooperation with the census that has been achieved.

Our concern is that trust in some of the bodies responsible for administrative data may not be as high as trust in the ONS. For example, it has recently been highlighted in Parliament that disabled people have low trust in the Department for Work and Pensions (DWP). It follows that, by moving away from the census, the ONS will implicitly be moving from ‘high trust’ statistical data sources, such as the census and ONS surveys, to lower trust data sources.

We acknowledge that the perceived trustworthiness of public bodies is an under-researched area: for example, the recent ONS Trust in Government (2022) survey only looks at the trust in a small range of highly-aggregated categories. While it shows a high level of trust in the NHS for example, according to the research Data-driven research and healthcare: public trust, data governance and the NHS (Kerasidou and Kerasidou, 2023) it appears that data sharing initiatives as developed by successive governments and implemented through the NHS have been met with public distrust. A much more detailed study of the trustworthiness of data collecting and sharing bodies needs to be undertaken before committing to administrative-based alternative to the census.

4(a): Refusals and community engagement

As noted above, there is also a major issue of people refusing to describe their ethnicity in the administrative data, which could well reflect lack of trust. Refusals could also be due to unwillingness to supply personal data that does not seem to the individual to be strictly necessary for administrative purposes.

Administrative bodies may also lack the resources for the kind of community engagement that the ONS is able to undertake for the census. This suggests that the quality of the administrative data may be lower than that of the census and may provide less valid data, especially about disadvantaged or at-risk groups where issues of trust are likely to be particularly important.

In contrast, the census provides a “moment” when everybody in the country participates in the same activity at the same time and “being counted in” can create a sense of belonging and inclusion (Killick, L., Duff, A., Hall, H., & Deakin, M., 2016).

4(b): Transparency

There is also the risk that the new statistical environment might be seen as less transparent, since it could potentially involve extensive modelling based on unverifiable assumptions as in the case of the DWP’s alternative claimant count series. It may also be worth noting that, since ONS does not have access to the data used in the modelling, the modelling has been undertaken by the DWP. The risk is that this might in turn undermine trust in the ONS’s own statistics and have the unintended consequence of reducing confidence in ONS’s survey programme. Even where ONS undertakes the modelling, full details of what particular model was chosen, and why, what assumptions have been made, and the sensitivity of the results to different assumptions are rarely communicated in a transparent way to stakeholders. The modelled data may therefore be opaque, and perhaps open to challenge.

In contrast to modelled administrative data, the census is highly transparent since people directly see the questionnaire and have the purpose and rationale clearly explained in the supporting material and engagement activities. The census also provides a practical demonstration of the principle that ‘everyone in society counts and is counted and no one is left behind.’ It thus performs an important inclusive function that is not, in our judgement, currently the case for many administrative datasets.