Social Capital Score

Shrinkage is applied by adjusting each LSOA’s indicator value towards its Local Authority average, with the amount of adjustment depending on how reliable the local data is (less reliable estimates are adjusted more). This is implemented using an empirical Bayesian approach (full formula provided in Appendix A of the report). This reduces random noise and prevents small-area results being overly driven by chance. While it introduces a small, intentional bias, this improves the overall accuracy and robustness of the social capital scores rather than distorting them. It trades a small amount of bias for a larger reduction in variance (noise).

All indicators are first aligned so that higher values consistently represent lower social capital. The indicators are then standardised, combined into dimensions, and aggregated into the final SCS – so a higher overall score reflects lower social capital (i.e. higher need). In short, areas with weaker outcomes across the underlying indicators accumulate higher scores, which is why “low social capital”
corresponds to a high social capital score.

The SCS is intentionally structured as a “need-based” measure, where higher scores indicate lower levels of social capital (i.e. greater need). This aligns with established indices like IMD and CNI, making results easier to interpret and compare. Also because the scores are exponentially transformed it draws out scores at the high needs areas so that you can better identify areas with challenges (the focal point for the analysis). Areas are then ranked so that rank 1 represents the lowest social capital (highest need), ensuring that the most disadvantaged areas are clearly identified.

While headline results may suggest higher scores in London, the methodology was explicitly designed to minimise systematic urban bias and ensure comparability across different settlement types. In particular, the SCS is constructed at LSOA level, the smallest feasible spatial scale for robust data, allowing it to capture within-area variation and avoid masking pockets of need that would be obscured
at coarser geographies, especially in rural areas where population dispersion is greater. The framework also explicitly accounts for structural differences between urban and rural contexts, including socio-economic heterogeneity and spatial distribution of populations, rather than assuming uniform conditions.

Importantly, the inclusion of asset-based indicators — such as access to civic assets and the concentration of third sector organisations introduces a counterbalancing effect, as these measures often highlight deficits in rural infrastructure and service provision. As a result, the SCS does not systematically privilege urban areas; instead, it identifies need wherever it occurs. This is reflected in the outputs, which show a number of rural and semi-rural areas such as parts of County Durham, Fenland, and Boston scoring poorly on social capital, demonstrating that the measure is sensitive to deprivation and social fragility beyond major urban centres.

A moderate correlation between Bridging and Linking social capital (0.5) is not unexpected and does not, in itself, indicate problematic conceptual overlap; rather, it reflects the reality that both domains are partly shaped by shared underlying structural factors such as socio-economic conditions, civic infrastructure, and patterns of participation. It is well established in composite indices that domains capturing related aspects of social and economic life will exhibit correlation, for example, several domains within the Indices of Multiple Deprivation show even stronger interrelationships without undermining their conceptual distinctiveness.

In this case, Bridging and Linking social capital remain analytically separable both theoretically and empirically: Bridging captures horizontal connections across diverse social groups (e.g. civic participation, volunteering, community assets), whereas Linking reflects vertical relationships with institutions and systems of power (e.g. civic engagement, voter turnout, access to formal structures). This distinction is strongly grounded in the social capital literature (e.g. Szreter & Woolcock, 2004; Putnam, 2000), and is reflected in the selection of indicators, which map onto different mechanisms and outcomes. The observed correlation therefore indicates complementary, not redundant, dimensions of social capital, each capturing a distinct but interrelated component of how communities connect internally, across groups, and with institutions.

Including population turnover within the Bonding social capital domain is consistent with a well-established body of literature linking residential stability to the formation of strong, trust-based local ties. Bonding social capital relies on repeated interaction, shared norms, and durable relationships within relatively homogeneous groups; high residential mobility disrupts these processes by shortening the time horizon over which relationships can develop and weakening opportunities for reciprocity and trust-building.

Empirical work on neighbourhood effects consistently finds that population churn is associated with lower levels of cohesion, trust, and informal social control for example, McCulloch (2003) shows that residential instability is negatively associated with neighbourhood social capital and contributes to social disorganisation, while earlier foundational work by Sampson, Raudenbush and Earls (1997) demonstrates that stable neighbourhoods are better able to sustain collective efficacy. Similarly, Putnam (2000) highlights residential stability as a key condition for the development of dense, inward-looking social networks, and subsequent studies (e.g. Forrest & Kearns, 2001; Kasarda & Janowitz, 1974) emphasise that high turnover undermines local attachment and neighbourly interaction. On this basis, population turnover operates as a meaningful proxy for the durability of local social ties: areas with higher churn are less conducive to the formation and maintenance of the close-knit, trust-based relationships that define bonding social capital, justifying its inclusion within this domain.

LSOAs are designed to have consistent population sizes, typically containing 400 to 1,200 households and 1,000 to 3,000 residents. They are the smallest statistical geography for which robust admin data from these sources can be produced. Ideally, when measuring social capital, we used the smallest geographic boundaries possible as those ties are hyper-local (e.g. neighbors, people you meet at park, etc.)

The maps display Social Capital scores grouped into quintiles (20% bands), showing how areas compare relative to others nationally. At LSOA level, the shading reflects all neighbourhoods across England, with each band representing 20% of areas, where darker colours indicate higher scores and therefore lower social capital (higher need).

The Local Authority maps follow the same principle but aggregate scores to LA level, so each band represents 20% of Local Authorities. The “hotspots” map focuses specifically on the 20% of Local Authorities with the lowest social capital (highest scores) and further divides this subset into five bands to highlight variation within the most disadvantaged areas.

Some degree of temporal misalignment between indicators is an inherent and recognised feature of composite indices that draw on multiple administrative and survey data sources, and reflects a deliberate methodological trade-off rather than a flaw. The primary objective of the SCS is to maximise coverage of robust, policy-relevant indicators at small-area level, using the most recent data available for each measure; in practice, enforcing a single uniform time point would significantly reduce the breadth and quality of indicators, particularly where high-quality data are collected infrequently.

This approach is well established in comparable frameworks, for example, the Indices of Deprivation 2019 combine data from different years, including Census 2011 alongside more recent administrative sources, on the basis that many underlying social and structural conditions change gradually and remain sufficiently stable for comparative purposes. While we acknowledge the limitation, the benefit of capturing a richer, multidimensional picture of social capital, grounded in the best available evidence, outweighs the loss of temporal alignment, and the methodology remains transparent about the timepoints used for each indicator.

The apparent compression of scores at the higher end of the distribution is a direct and intentional consequence of the exponential transformation applied during standardisation. This transformation is designed to stretch out the upper tail of the distribution, where the highest levels of need (i.e. lowest social capital) are concentrated, while compressing the lower end, thereby improving discrimination between the most disadvantaged areas and reducing “cancellation effects” across domains.

In practice, this means that differences among high social capital areas
(low-need areas) are deliberately deemphasised, as the primary analytical purpose of the measure is to identify and prioritise areas with the greatest deficits in social capital. As such, the absolute score values are less meaningful in isolation; the measure is fundamentally a relative measure, and it is the ranking of areas that should be used for interpretation and comparison. This approach is consistent with established methods used in indices such as the Indices of Deprivation, where transformed scores are intended to support relative ordering
rather than precise interval-scale interpretation.

While these forms of internal validation were not the primary focus of the current methodology, which instead emphasises transparency, theoretical grounding, and the use of robust small-area indicators we recognise their value and would be open to exploring them as part of future development. Indeed, we did initially consider using Maximum Likelihood Factor analysis to determine weights.

In terms of external validation, initial steps have been undertaken: the SCS has been compared with related measures such as the Indices of Deprivation and the Community Needs Index, demonstrating expected positive relationships and providing evidence that the measure is capturing meaningful underlying social and structural conditions. Further work could extend this by examining associations with additional outcomes (e.g. health, crime, or resilience metrics), but the existing correlations already provide a degree of reassurance as to the
construct validity of the framework

A relatively strong correlation with deprivation (r = 0.672) is not unexpected, but it does not imply that the Social Capital Score is simply duplicating existing deprivation measures. The two constructs are conceptually distinct: deprivation indices are designed to capture material disadvantage and access to economic resources (e.g. income, employment, housing), whereas this framework is explicitly
focused on the social infrastructure of communities, relationships, trust, civic participation, and connections to institutions.

Consistent with this distinction, the indicators used in the Social Capital Score are largely non-overlapping with those in deprivation indices, drawing instead on measures of neighbourhood relationships, civic engagement, social trust, and community assets. The observed correlation is therefore best understood as reflecting shared underlying structural conditions, areas facing economic disadvantage are also more likely to experience weaker social networks and lower civic capacity, rather than conceptual redundancy. Importantly, the relationship is far from perfect, and there remains substantial variation between the measures, indicating that social capital captures additional dimensions of community resilience and cohesion that are not directly measured by deprivation indices.

We recognise the limitations associated with drawing on multiple indicators derived from the same underlying survey source. The use of the Community Life Survey reflects a broader constraint in measuring social capital at the small-area level: the relative scarcity of granular data capturing concepts such as trust, relationships, and civic participation. While the indicators derived from this source do share a common methodology and may therefore introduce correlated measurement error, they have been included because they provide some of the most direct and policy-relevant proxies for these otherwise hard-to-measure aspects of social capital.

Importantly, the framework does not rely solely on these survey-based measures; they are complemented by a wider set of administrative and
behavioural indicators (e.g. crime, residential mobility, civic assets, third sector presence), which help to triangulate the construct and reduce over-reliance on any single data source. Nonetheless, we acknowledge this as an area for improvement and would seek to incorporate additional or alternative data sources in future iterations to strengthen robustness and independence across indicators.

While we recognise that some indicators, such as crime and residential mobility, can be interpreted as outcomes of low social capital, strong evidence also supports their role as valid proxy measures within the construct, particularly in area-based frameworks. In the case of residential mobility, a well-established body of research (including McCulloch, 2003) demonstrates that high population turnover disrupts the formation of stable, trust-based relationships, weakening neighbourhood cohesion and reducing opportunities for repeated social interaction — both of which are central to bonding social capital. As such, mobility is not only an outcome but also a key mechanism through which social capital is eroded.

Similarly, crime has been widely used in empirical studies as an indicator of social capital and collective efficacy; for example, Irfan, M.,et al. Mapping social capital across Wales (UK) using secondary data and spatial analysis. SN Soc Sci 3, 56 (2023) include crime within their operationalisation of social capital, reflecting the well-documented relationship between lower levels of trust, weaker informal social control, and higher crime rates (building on foundational work such as Sampson et al., 1997). In this context, these indicators are best understood as part of a mutually reinforcing system, where social capital both shapes and is shaped by local conditions. Their inclusion therefore reflects a pragmatic and literature-supported approach to capturing the observable manifestations and structural correlates of social capital at neighbourhood level, rather than implying simple one-directional causality.

It is reasonable to acknowledge that Facebook-derived metrics may introduce demographic and platform-usage biases, for example, differences in uptake by age, socio-economic group, or digital engagement. However, there is also a strong and growing body of evidence suggesting that these data can provide highly robust and policy-relevant insights when used appropriately and at scale. In particular, the Behavioural Insights Team and associated research underpinning the UK Social Capital dataset highlight the unprecedented scale and coverage of these data: analyses draw on 20 million UK residents covering a substantial proportion of the adult population (e.g. around 58% of UK adults aged 25–64) https://www.bi.team/wp-content/uploads/2025/03/Social-Capital-in-the-United-Kingdom-Research-summary.docx.pdf. This scale significantly reduces random error and enables highly granular, small-area estimates that are not feasible using traditional survey methods.

Moreover, these datasets are not used naively; they are typically processed using privacy-preserving and statistical techniques (e.g. aggregation and noise injection) specifically designed to maintain reliability while mitigating bias. https://data.humdata.org/dataset/uk-social-capital-atlas?

On this basis, while we acknowledge the potential for demographic bias and treat these indicators with appropriate caution, their inclusion is justified by their exceptional scale, granularity, and ability to capture real patterns of social connection. They are also used alongside a broader set of administrative and survey-based indicators, ensuring that the overall framework does not rely on any single data source and that potential biases are mitigated through triangulation.

This is a reasonable suggestion, one we explored during the development process. Correlation analysis was undertaken across the core indicators. It found a generally high degree of correlation within each of the three conceptual domains (Bonding, Bridging and Linking social capital), reassuring us that indicators grouped within each domain captured related underlying constructs.

Importantly, the analysis suggested stronger relationships within domains than between them, which supports the conceptual validity of retaining the three-part framework. Had indicators shown weak relationships within domains despite conceptual similarity, we would have considered restructuring the framework, for example through the creation of sub-domains or by reallocating indicators to different dimensions where appropriate.

At the same time, we did not seek perfect correlation between indicators, as social capital is inherently multidimensional and different indicators are intended to capture distinct manifestations of community cohesion, participation and institutional connection. Excessively high correlations across all indicators could instead suggest redundancy and duplication rather than a richer multidimensional framework.

We would be happy to include additional summary correlation analysis in future iterations of the report to provide greater transparency regarding the index’s internal structure.

The decision for higher scores to represent lower social capital was primarily driven by the use of exponential transformation during the standardisation process and by the policy purpose of the measure.

For a non-technical audience, exponential transformation can be thought of as a way of “stretching out” the areas with the greatest levels of need, while compressing areas performing relatively well. This makes it easier to distinguish between neighbourhoods experiencing the weakest social capital and prevents poor performance in one dimension from being completely offset by stronger performance elsewhere.

For example, without this approach, an area with very weak neighbourhood trust but relatively strong civic participation could appear “average” overall because the two effects cancel each other out mathematically. The exponential transformation reduces this cancellation effect by giving greater emphasis to areas experiencing
particularly poor outcomes on one or more dimensions.

As a result, the methodology intentionally focuses analytical attention on neighbourhoods with the weakest social capital, because these are the areas most likely to require intervention, investment or policy support. In this context, having higher scores correspond to poorer outcomes aligns the measure with its practical policy use: areas with the highest scores are those facing the greatest challenges.

Importantly, the index is fundamentally a relative ranking measure rather than an absolute scale. The precise numerical values are therefore less important than the relative ordering of areas and the identification of neighbourhoods experiencing comparatively lower levels of social capital.

The Social Capital Score has been designed to provide the most comprehensive and robust measure of social capital possible at neighbourhood level. In developing the framework, we prioritised the inclusion of indicators that are conceptually relevant to the dimensions being measured and sufficiently robust to support reliable comparisons at small-area geography, rather than restricting the measure solely to the most recent datasets available.

This inevitably results in some variation in the reference periods of the underlying indicators, as many of the datasets that capture important aspects of social capital—such as neighbourhood relationships, trust, civic engagement and socio-economic characteristics—are collected at different frequencies. Restricting the framework only to datasets from a single time period would substantially reduce the breadth of evidence available and omit a number of important dimensions of social capital.

This approach is consistent with established practice in the construction of composite indices. For example, the English Indices of Deprivation 2019 combined indicators from multiple time periods and included data from the 2011 Census alongside more recent administrative sources. The rationale is that many underlying social and structural characteristics change relatively slowly over time and continue to provide valuable evidence of local conditions even where more recent updates are not available.

We recognise the limitations associated with temporal misalignment and have therefore been transparent about the date and source of each indicator included in the measure. As new datasets become available, the framework can be updated and refined. However, we believe that the benefits of drawing on the broadest possible evidence base outweigh the disadvantages of excluding theoretically important indicators solely on the basis of their age.

Although the two measures are related, they are designed to measure different concepts.

The Community Needs Index (CNI) focuses on the social infrastructure available within a place. It measures factors such as the presence of community and civic assets, connectivity to services and opportunities, and levels of community participation and engagement. The CNI was originally developed to identify communities that may lack the physical, civic and institutional resources needed to support local action and resilience.

The Social Capital Score focuses on the strength of social relationships within communities. It measures factors such as trust, neighbourhood cohesion, civic participation, social networks, community stability and the connections between residents, organisations and institutions.

In simple terms, the Community Needs Index is primarily concerned with whether communities have access to the assets, infrastructure and opportunities that support community life, whereas the Social Capital Score is concerned with the strength of the social bonds, relationships and civic connections that exist between people.

The two measures should therefore be viewed as complementary rather than interchangeable. Strong social infrastructure can help create the conditions for social capital to develop, while high levels of social capital can help communities make effective use of local assets and opportunities. Used together, the measures provide a richer understanding of both the resources available to a community and the strength of the social relationships that underpin community wellbeing and resilience.

The Social Capital Score is a relative measure of social capital. It is designed to compare neighbourhoods with one another rather than to provide an absolute measure of how much social capital exists in a place.

The most important outputs are therefore the rankings and relative positions of areas, rather than the numerical scores themselves. A neighbourhood’s score indicates how it performs relative to other neighbourhoods in England on the indicators included in the Index. The scores should not be interpreted as representing a fixed quantity of social capital, nor should differences between scores be interpreted as representing proportionate differences in social capital.

To support policy and targeting decisions, the Social Capital Score has been constructed so that higher scores indicate lower levels of social capital and higher levels of need. Conversely, lower scores indicate stronger social capital. This means that a neighbourhood ranked 1 has the lowest level of social capital (highest need) relative to all other neighbourhoods in England, while neighbourhoods with the largest rank values have the strongest levels of social capital.

This approach was adopted deliberately because the primary purpose of the measure is to identify communities that may benefit from additional support, investment or intervention. Presenting the Index in this way enables users to focus directly on the neighbourhoods experiencing the greatest social capital challenges.

Users are therefore encouraged to interpret the results in terms of relative position, for example whether an area falls within the most disadvantaged 10%, 20% or 50% of neighbourhoods nationally, rather than focusing on the absolute value of the score itself. As with other widely used composite measures, such as the Indices of Deprivation, the rankings provide the most robust basis for comparing areas and identifying those with the greatest relative need.