Social exclusion has been discussed in many disciplines for decades, yet previous efforts to measure this exclusion have not captured the multidimensional complexities of social exclusion. We developed a five-domain model of social exclusion: material, relational, political, digital, and structural, based on the academic literature and expert consultation. We used items from a longitudinal household survey to parameterise this model, creating domain-specific and overall social exclusion scores.
We showed how social exclusion varied by age, ethnicity, and region, with high levels of social exclusion experienced in the youngest age groups, minority ethnic groups, and those living in London or the north of England. We found striking differences by domain. Taken together, these results demonstrate the utility of a multidimensional measure to provide overall and domain-specific estimates of social exclusion.
When comparing the estimates from 2009/10 to 2018/19, we found stability for social exclusion and its domains by age, showing decreasing levels of social exclusion with increasing age. Further, levels of social exclusion were higher in ethnically minoritised groups compared to the White British group at both time points, which appeared to be driven by material, relational, and structural exclusion. We noted interesting changes by geographic region between the two time points. In 2009/10, four regions had very low levels of social exclusion: South West, East Midlands, South East, and Yorkshire & the Humber. In contrast, the results from 2018/19 showed clearer evidence of a North-South divide in exclusion scores, where regions north of the Severn-Wash line had high levels of social exclusion (North East, Yorkshire & the Humber, North West, East Midlands), while Southern regions had low levels. While this may indicate growing regional disparities in social exclusion, more work is needed to explore the sensitivity of this measurement approach and its stability over time.
There is some debate in the literature about if someone would be considered socially excluded if they experience exclusion on a single domain. Some research suggests that individuals must face exclusion in multiple domains to be considered excluded, while others maintain that low participation and exclusion in any domain is sufficient to being classified as excluded [6, 19]. In our measurement approach, we used standardised sums for each domain to create an overall measure of social exclusion, describing those with scores above the mean as experiencing social exclusion, to varying degrees. However, as this is a mean score based on domain scores, groups may experience high levels of exclusion in one domain and low levels of exclusion in another, so would receive a social exclusion score near the mean. Rather than focusing on a cut-off point, we suggest that these measures should be viewed as a continuum, which allowed a more nuanced understanding of social exclusion. For example, an individual may experience high exclusion in a single domain, while others may experience moderate levels of exclusion across multiple domains. These differing circumstances would require different policy and intervention responses.
While it is useful to have an overall measure of social exclusion, the summary score masks the heterogeneity in the domains. Being able to disaggregate social exclusion into its parts can be useful to inform public health interventions. For example, while those 65 or older had the lowest overall social exclusion scores at both time points, the high level of digital exclusion indicates that this group may benefit from interventions with support their digital inclusion. The youngest age group had the highest level of social exclusion across the age groups, which was driven by high levels of relational exclusion and elevated scores in the structural and material domains. This trans-domain elevation of social exclusion may require a multi-faceted public health intervention which supports young people in developing positive social relationships, supports their inclusion in employment and access to adequate housing, and addresses structural barriers. This research highlights different drivers of social exclusion in population groups. This may inform targeted public health responses to reducing exclusion and supporting inclusion of vulnerable population groups.
Comparison to previous literature: The five domains in our social exclusion score build on previous work measuring social exclusion, most notably the Social Exclusion Index for Health Surveys (SEI-HS) [15]. SEI-HS’s material deprivation domain was like our material exclusion domain, the social participation domain was analogous to our relational exclusion domain, and measures within “lack of normative integration” were like measures in our political exclusion domain. We expanded political exclusion to include political participation and engagement in addition to contributing social causes and acting for collective good. We included digital and structural exclusion, which were not measured in the SEI-HS, but were supported by our literature review of social exclusion and consultations with our expert advisory panel. While there was overlap in our measurement approaches, the SEI-HS was derived based on measures in the Netherlands Public Health Monitor, which included many items which are not routinely collected in the UK, including feeling cut off from people, feeling rejected, having people who understand you, receipt of medical and dental treatment, and giving to good causes. Future research could investigate how the SEI-HS compares to our measurement approach if a suitable dataset is identified.
The summary measure of social exclusion was able to highlight patterns across key demographic factors and our findings align with previous literature. The overall score, for example, provided evidence of the North-South divide, a geographic phenomenon observed in previous research where regions north of the Severn-Wash line experience higher levels of deprivation, higher levels of premature mortality, and other poor health outcomes compared to than southern regions [20, 21]. With the notable exception of London, southern regions had lower social exclusion than the rest of England. These differences were particularly apparent in the material and structural domains. The high level of social exclusion in London aligned with previous findings in older adults which showed that there was elevated social exclusion across multiple domains.22 In contrast to previous findings [22, 23], we found a strong inverse relationship between age and social exclusion, with greater social exclusion at younger ages.
Limitations: While USoc includes many measures, the available variables did not fully capture the theoretical concepts for each domain. For example, there were few variables available for the digital domain and were not adequate to capture variations in access, resources, knowledge, and skills. In 2018/19, most households reported having a computer and internet connection, but the persistent digital divide in England suggests that access alone is not sufficient for digital inclusion [24, 25]. An important construct within the structural domain was discrimination, however, discrimination questions were only asked of less than 5% of the sample so had to be excluded. In the political domain, we wanted to estimate taking collective action but were limited to using environmental habits as a proxy, which misses other ways an individual may take action. Further, while we acknowledge that social exclusion is shaped by broad structural factors, including social and cultural norms, government policies, wider economic conditions, and global events [26, 27], these were not captured in USoc and our analysis was conducted primarily on individual- and household-level variables.
This research focused on the experience of social exclusion which was non-voluntary and linked to broader circumstances and systems which prevented full participation in society which were beyond the control of an individual. Some individuals choose not to participate in aspects of society, like choosing not to be involved in political processes or choosing to eschew technology due to personal beliefs and preferences. We were not able to determine if participants chose to not participate in various aspects of life (voluntary) or if they were socially excluded (non-voluntary).
Strengths: Despite limitations in the measures, USoc includes extensive measures which made it possible to explore social exclusion in detail, which might not be possible in other studies. This approach captures the multidimensional nature of social exclusion. Our domains overlap with and extend previous research, by adding a digital exclusion domain, which has not appeared in previous indices and has emerged as an increasingly relevant aspect of exclusion.
A further strength of this approach is that it outlines a method which can be used to generate comparable social exclusion scores which can be used in population surveys, even if the exact measures are not included. This measurement approach may improve the generalisability of this measurement approach, as future waves of this survey and other population surveys may use different measures of income or education, so by including the available measures which best estimate the constituent components of each domain, we would expect broadly similar results. This represents a pragmatic use of existing data, as population surveys have been developed to capture a broad picture of the health of the population and often do not have social exclusion as their primary focus. This allows us to estimate social exclusion using existing data. The flexible and pragmatic approach to estimating the domains permits longitudinal follow-up in the context of changing survey design. It also allows for the estimation of regional levels of social exclusion, which was not possible in smaller surveys.