Case Study D: Joseph Rowntree Foundation (JRF)

Taha Bokhari is a Lead Analyst at the Joseph Rowntree Foundation (JRF). He leads on the production of “UK Poverty”, a flagship report that helps to anchor our collective understanding of poverty trends using the most recent data available. It presents snapshots and trends of poverty rates broken down by a range of demographic and socio-economic characteristics over time. 

Following feedback from a network of faith-based stakeholders, JRF sought to include religion as one of their breakdowns in UK Poverty 2026. Data about religion is collected in the Family Resources Survey (FRS), which underpins HBAI. JRF regularly access this data via the UK Data Service (UKDS) each year when the produce the report. 

However, when they attempted to incorporate religion, they discovered that the relevant variables were available only under UKDS Special Licence, which requires additional application process involving detailed justification of purpose, skills, and data handling procedures. The application process created two noteworthy points of frustration: 

1. Approval was expected to take three to four months, which was incompatible with the timeline for the report’s publication timeline. 

2. The process was described as opaque, with no clearly designated contact point responsible for responding to queries. It was ultimately unclear why data access had been restricted in this way for the HBAI data when it is publicly available in similar publications for Scotland and Northern Ireland.  

As a workaround, Taha pivoted to using Understanding Society to produce indicative poverty rates by religion. However, this introduced the need for various caveats on the results. Poverty rates in Understanding Society are not calculated in exactly the same way as in HBAI, and the dataset does not include all components required to replicate the HBAI methodology (for example, some housing benefit information available in HBAI is not available in Understanding Society). The end result was added work for the JRF team and analysis of religion that was more modest than they had hoped. 

For future publications, they are considering using Freedom of Information requests for the religion data, since it can produce results faster than getting approval for secure access, although it is also unreliable in some ways and can receive responses of varying quality. From the producer perspective, that is likely to be a suboptimal solution as well, since they would need to invest time and resources fielding one or multiple requests. 

Moreover, the lack of support throughout the process of getting access to the special license data added unnecessary friction for the research team. Indeed, it is worth noting that engagement with Understanding Society was considered significantly easier—they have comparatively responsive user-support services that can helpfully respond to queries about the data and how to access it.  

This experience illustrates a broader point. When certain characteristics are missing or inadequately captured in FRS/HBAI, users that want to continue their analyses may turn to alternative surveys such as the Labour Force Survey or Understanding Society. However, analysis of poverty using these datasets requires recreating poverty indicators from scratch, taking into account differences in income measurement and benefit variables. This is resource-intensive and methodologically imperfect. 

Taha noted that many alternative surveys already provide extensive derived variables, yet do not routinely include a derived poverty flag. The absence of such a variable increases barriers to analysis, particularly for smaller organisations without the internal capacity to construct poverty measures themselves. Without derived variables included by default in household survey data, organisations looking to study poverty will generally be reliant on data from the FRS. This means that gaps in the FRS will ultimately constrain those organisations, too.  

A relevant example of where this matters could include formal care use. Taha noted that he would like to model the relationship between poverty status and receipt of formal care, controlling for other characteristics. However, like most household surveys, the FRS does not contain a sufficiently robust variable identifying individuals in formal care, because people in institutional settings tend to be excluded from their sampling frames altogether.  

This makes it difficult to assess poverty among those receiving formal care or to examine how poverty and care needs interact. If another survey was able to capture people in formal care settings better and include a derived poverty flag, then poverty analysis in this context would open up to a diverse range of analysts, including those at JRF and smaller organisations. 

In summary, this case points to a few possible improvements to the current system that may be relatively easy to action. These include: 

• Investment into better user-support mechanisms, including clearly designated points of contact and people responsible for handling queries. 
• Greater transparency around how and why access restrictions are made. 
• Built-in construction of poverty flags in major household surveys to make analysis of poverty more accessible to a wider range of users.