Introduction: Voluntary action and the local context

The role and impact of voluntary action in the United Kingdom is once again on the political and scholarly agenda. The COVID-19 pandemic and the subsequent cost of living crisis highlighted the contribution of voluntary organizations to their communities and gave rise to optimism that those contributions might be expanded (Kruger, Reference Kruger2020). However, there are long-standing concerns that the uneven distribution of voluntary organizations across local areas would limit the ability of voluntary action to reach all communities (Mohan, Reference Mohan and Morvaridi2015; Wolfenden, Reference Wolfenden1978). As John Stuart Mill (Reference Mill and Nathanson1848: 299) pithily surmised: “Charity almost always does too much or too little. It lavishes its bounty in one place, and leaves people to starve in another.”

The uneven distribution of voluntary organizations matters because they provide direct community-level benefits, including the diverse and direct contributions they make to quality of life, supporting disadvantaged or excluded groups, and providing outlets for prosocial activities. There are also contentions that latent benefits for community functioning arise from a strong voluntary sector presence, primarily through the formation of social capital, with the beneficial consequences that are said to flow from it (Sharkey et al., Reference Sharkey, Torrats-Espinosa and Takyar2017). Recognition of these contributions has meant that there is also practitioner interest from policymakers and stakeholders in the importance of the spatial distribution of voluntary organizations because of the contribution they might make to the reduction of regional and local disparities in prosperity and well-being. Drawing on that context, this article is motivated by the need to interleave recent and historical snapshots of voluntary action in a coherent, rigorous manner. Using longitudinal data on charitable organizations across six census periods (1971–2021), we estimate growth trajectories in the number of charities per capita for local authorities across England and Wales. We assess whether the shape of these trajectories is associated with the demographic mix of an area’s residents (e.g., levels of education, or age and ethnic composition) and of its location, including its level of material deprivation, geographic region, and degree of rurality. We examine whether there is substantial variation in the level and shape of these trajectories across local authorities or whether there is a “national” trajectory for the development of voluntary organizations in the United Kingdom. Our findings suggest that the latter is the case: charity density grew rapidly from 1971 to 2001, followed by a noticeable plateau and decline to 2021, with this trend consistent across almost all local authorities. In addition, we reveal enduring and widening equalities in the level of charity density across local authorities, and minimal relative movement of areas up or down the distribution. Our findings suggest that spatial inequalities in the distribution of charitable organizations are substantial and persistent, which raises important questions about how far these inequalities might be modified, whether through public policy or through private initiative.

We first describe theoretical explanations for unevenness in the distribution of voluntary organizations across geographic contexts, followed by a review of important empirical work on this topic. We then consider issues of data and method before presenting the main empirical findings. We situate these findings in the context of prior theoretical and empirical contributions before concluding with a consideration of the implications for policy and practice, against the background of public policy concerns about enduring spatially concentrated disadvantage in the United Kingdom.

Theoretical foundations: Variation in voluntary activity and its implications

Voluntary organizations, of which charities are a significant subset, are often defined using the Structural-Operational definition developed by Lester Salamon and colleagues: they are formally constituted, nonprofit distributing, autonomous, and therefore not controlled by state entities, and they benefit from significant voluntary contributions of time and money from their unpaid board members (trustees), volunteers, and donors (Salamon & Anheier, Reference Salamon and Anheier1992). In the United Kingdom, registered charities form the largest component of the organizational base of the voluntary sector. Their emergence of a voluntary sector, at the national and local levels, is often explained with reference to classical demand and supply-side theories. The former posit that the government or market undersupplies public goods and services in an area and thus there is latent demand from the populace that voluntary organizations can satisfy—especially when it comes to services for issues that affect minority groups (Hansmann, Reference Hansmann, Powell and Steinberg1987; Weisbrod, Reference Weisbrod and Phelps1975). These demand-side theories are framed as market or government failures. Supply-side theories contend that organizational formation is contingent on the availability of sufficient “enabling resources” in an area, such as a pool of willing and able volunteers and founders, or diverse and adequate funding sources (James, Reference James and Powell1987; Musick & Wilson, Reference Musick and Wilson2008). Interdependence theory contends there is a symbiotic relationship between government and the voluntary sector: the former recognizes and leverages the value of these organizations in addressing social needs, whereas the latter depends on government for sufficient resources and legitimacy (Mayer, Reference Mayer2024; Salamon, Reference Salamon1987). The supply-side perspective thus provides a strong basis for expecting unevenness in the distribution of voluntary organizations according to compositional differences in socioeconomic resources, especially the availability of government funding and donations (Logan & Molotch, Reference Logan and Molotch1987) and volunteers/social entrepreneurs (James, Reference James and Powell1987). For example, deprived areas may struggle to generate sufficient financial capital to sustain a thriving voluntary sector, while also lacking a pool of individuals with the capacity to engage consistently and intensively in voluntary work.

These frameworks are complemented by deficiencies in the operational form and logic of voluntarism. Voluntary failure theory (Salamon, Reference Salamon1987) describes two of these as “resource insufficiency” (the inability of a voluntary organization to generate and sustain sufficient operating resources) and “philanthropic particularism” (the disjuncture between the preferences of voluntary organization founders/trustees and “priority” areas of need). However, rather than considering these as flaws, Toepler (Reference Toepler2023) argues that they represent drivers for increased collaboration between the voluntary sector and government (interdependence theory). For example, resource insufficiency can be addressed by targeted government funding and policy programs (e.g., initiatives aimed at the regeneration of areas of concentrated disadvantage), while philanthropic particularism can be tempered by government stipulating fair access to voluntary goods and services (e.g., through charitable or equalities law).

Finally, ecology theories posit that the organizational environment in which voluntary organizations operate shapes the size and sustainability of local voluntary sectors (i.e., their density). These theories argue that organizations are exposed to environmental selection pressure in the form of competition for resources and search for legitimacy, which in turn influences the number of organizations that can be “carried” or present in a field or community (Carroll & Hannan, Reference Carroll and Hannan2000; Hannan & Freeman, Reference Hannan and Freeman1989). For example, Mayer (Reference Mayer2025) demonstrated how nonprofit foundation and dissolution rates change in response to shifts in density of these organizations within U.S. counties. Organizational ecology theory potentially explains why some areas have higher levels and growth rates of voluntary organizations than others.

If inequalities are to be expected, do they matter? There are important consequences for voluntary organizations themselves (Crubaugh, Reference Crubaugh2021; Mayer, Reference Mayer2023, Reference Mayer2025): their prospects of survival are influenced by geographical context. In a comparison of more disadvantaged versus more prosperous areas, rates of charity formation are lower, and the risk of charity dissolution is higher, in more deprived areas (Clifford, Reference Clifford2018). Therefore, inequalities in the distribution of these organizations are enduring and sustained through reinforcing processes of foundations and dissolutions that are shaped by the local context. Inequalities affect the ability of local populations to access goods and services provided by voluntary organizations. Voluntary organizations contribute directly to the “mixed economy of welfare” (Billis & Glennerster, Reference Billis and Glennerster1998), they serve marginalized groups whose needs are not well met by mainstream programs (Musick & Wilson, Reference Musick and Wilson2008), and they offer opportunities for volunteering and social entrepreneurship (Ben-Ner & Van Hoomissen, Reference Ben-Ner and Van Hoomissen1992).

A strong presence of such organizations also provides latent benefits to communities. First, it influences citizen engagement in voluntary action—there are small, but statistically significant, positive associations between the distribution of nonprofit organizations and the likelihood of volunteering (Deri-Armstrong et al., Reference Deri-Armstrong, Devlin and Seifi2023; Mohan & Bennett, Reference Mohan and Bennett2019; Rotolo & Wilson, Reference Rotolo and Wilson2012). Second, in Sampson’s (Reference Sampson2012) major study of neighborhood development in Chicago, variations in the distribution of nonprofit organizations were seen as part of a community’s infrastructure, linking individuals together in unintended ways that enhanced tasks oriented toward the public good, such as mobilizing claims for resources from city government (Sampson, Reference Sampson and Demeulenaere2011). Third, the stability of the distribution of nonprofit organizations over time was consistent with Sampson’s (Reference Sampson2012) wider arguments that neighborhoods are durable features of the social fabric. Thus, dense networks of voluntary organizations help create spaces for political association and activism, shaping access to government support and services, and encouraging community responses to social problems, thereby strengthening collective efficacy (Billis & Glennerster, Reference Billis and Glennerster1998; Brandtner & Dunning, Reference Brandtner, Dunning, Powell and Bromley2020; Clifford, Reference Clifford2021; Crubaugh, Reference Crubaugh2021; Sharkey et al., Reference Sharkey, Torrats-Espinosa and Takyar2017). In summary, there is a growing appreciation of the value of adopting a spatiotemporal perspective when studying voluntary organizations and their environmental conditions (Macindoe & Oakley, Reference Macindoe and Oakley2023; Mayer, Reference Mayer2023, Reference Mayer2024).

Empirical evidence: the extent and predictors of variation in voluntary activity

These theoretical predictions have borne out in studies of voluntary sectors in many countries. In the United Kingdom, there are sizeable and enduring differences in the numbers of charitable organizations between and within local government areas (Clifford, Reference Clifford2012, Reference Clifford2018; McDonnell et al., Reference McDonnell, Mohan and Norman2020), with similar findings derived from studies in the United States and South America (da Costa, Reference da Costa2016; Dalvi et al., Reference Dalvi, Mainardes and Sepulcri2024; Mayer, Reference Mayer2023; Peck, Reference Peck2008; Wo, Reference Wo2018). What drives these observed inequalities? The level of social need or material deprivation in an area is predicted to have a substantial effect on the number of voluntary organizations present or operating there. The evidence is mixed but tends toward supporting supply-side theories of emergence: on average, there are more organizations present in affluent than deprived areas. This relationship is often nonlinear, with a number of studies revealing a relationship with a convex functional form between deprivation/need and voluntary organization density. In other words, the most deprived areas have higher levels of organizations than their slightly less deprived counterparts, but lower levels compared with the most affluent areas (Clifford, Reference Clifford2018; McDonnell et al., Reference McDonnell, Mohan and Norman2020). Studies that find that density increases as deprivation worsens tend to be focused on specific subsectors, such as antipoverty nonprofits (Yan et al., Reference Yan, Guo and Paarlberg2014; see also Peck, Reference Peck2008). Other studies point to more complex and ambiguous connections between voluntary organization density and social need. For example, indicators of need (unemployment and poverty), ethnic diversity, and income inequality were not always associated with nonprofit density in a consistent manner, and some nonlinear relationships were also found (da Costa, Reference da Costa2016; Van Puyvelde & Brown, Reference Van Puyvelde and Brown2016; Wo, Reference Wo2018; Wu, Reference Wu2021). Zhulke (Reference Zhulke2022) argues that the mixed evidence on the association between demand and supply-side factors and nonprofit density reflects a deficiency in theoretical and methodological framing: studies either do not test the association of both types of factors in the same study, or, when they do, they fail to link or interact these factors.

This article makes a distinctive contribution by estimating and explaining the growth process that produces uneven patterns in voluntary organization density across local areas. By combining longitudinal data and latent growth-curve models, we show, for the first time, how the density of charitable organizations has developed over a 50-year period in England and Wales overall and for each local authority specifically. We achieve this by specifying two research questions. First, do local authorities differ significantly in their charity density trajectories over time? Second, what factors determine the level and trend in these trajectories? There are very few analyses of how density changes over time and whether there is spatial variation in this change, yet such work can reveal local contexts where growth in charity density is particularly strong or weak, and thus whether inequalities have worsened or improved over time.

Data and method

We focus on charitable organizations that, at some point in their history, have been formally registered with the Charity Commission of England and Wales (“The Commission”). Charities are a subset of the voluntary sector that—in addition to being formal, nonprofit distributing, autonomous, and benefitting from voluntary contributions (Salamon & Anheier, Reference Salamon and Anheier1992)—must operate for public benefit to possess charity status. We use a recent version of the Charity Register that provides a census of charities registered in England and Wales between 1961 and 2021 inclusive (N = 336,193). The register contains information on year of registration and deregistration, geographic scale of activity, and registered office address and postcode. Information on field of activity (e.g., Social Services and Education) is drawn from the UK Charity Activity Tags project and linked deterministically via registered charity number. Using the National Statistics Postcode Lookup (May 2022) file, we identify the local authority a charity’s postcode is in and create summary measures of charitable activity for each local authority and registration year: the number of charities present at the beginning of each year (a); the number of new charities registered that year (b); the number of existing charities de-registered that year (c); and finally the total number of charities present at the end of the year (d = a + (b − c)). We exclude observations for the City of London (which is one of the 33 local authorities in the greater London area). This local authority is geographically (1.12 square miles) and demographically (population size <10 k) small but has several thousand registered charities: many of these organizations do not operate in the City of London and are in many cases long-established grant-making trusts registered through solicitors’ offices. Thus, their presence is driven by different foundation and dissolution processes than in other local authorities. Moreover, the City of London’s distinctive history (as an administrative unit strongly associated with the financial district with which it shares a name) and organizational structure (it does not share the same functions as all other local authorities and, highly unusually, has its own police force) mean that it is defensible to exclude it from this analysis.

Sociodemographic data for each local authority have been obtained from the six censuses from 1971 to 2021 in England and Wales. To account for boundary changes for administrative areas over time, the data are adjusted to a common set of spatial units (Norman et al., Reference Norman, Rees and Boyle2003)—the 2021 definition of the local authorities. Previous repositories of consistent variables over time in terms of both definition and geography have been updated to 2021 and extended to include additional variables than were previously available (Lloyd et al., Reference Lloyd, Catney, Noble, McLennan and Norman2022; Norman, Reference Norman2016, Reference Norman2017; Norman et al., Reference Norman, Lloyd, McLennan, Ferguson and Catney2024).

The data on charities and local authorities are linked using local authority geocodes. The result is a balanced panel containing 1,980 observations across six census periods for 330 local authorities in England and Wales. We then calculate our outcome of interest, charity density, as the number of charities per 1,000 residents for a given local authority and census year. Our key covariates are the level of deprivation in a local authority, measured using the Townsend Score, the degree of rurality, UK region, ethnic composition, and level of education. The Townsend Score is a long-established composite measure comprising four variables that capture different aspects of disadvantage, measured for geographical areas. They are the proportions experiencing unemployment (signaling lack of material resources and insecurity), overcrowding (an indicator of living conditions), lack of owner-occupied accommodation (a proxy for household wealth), and lack of car ownership (a proxy indicator of income). These have been available consistently over time in the UK census since 1971. The index is a Z-score, in which where a value of zero represents a local authority with a mean level of material deprivation, above zero identifies more deprived areas relative to mean deprivation, and below zero identifies more affluent areas relative to mean deprivation (cf. McDonnell et al., Reference McDonnell, Mohan and Norman2020). Rurality is operationalized as an ordinal categorical measure of the number of persons per hectare (“Most Urban,” “Very Urban,” “Urban,” “Rural,” and “Most Rural”)—previous studies have shown that rural/urban differences in density are expected, and including this measure alleviates some of the confounding that can occur between density and material deprivation (Clifford, Reference Clifford2012; da Costa, Reference da Costa2016); UK region, measured using the International Territorial Levels classification (“North East,” “North West,” “Yorkshire and The Humber,” “East Midlands,” “West Midlands,” “East of England,” “London,” “South East,” “South West,” and “Wales”); ethnic composition, measured as the percentage of non-White individuals (non-British Isles born in 1971 and 1981, and non-White ethnicity 1991 onwards); and level of education, measured as the percentage of individuals with the highest level of educational qualification (i.e., degree). Population structures, such as ethnic, educational, and educational diversity, have consistently been found to be associated with levels of voluntary or civic activity (Borkowska et al., Reference Borkowska, Kawalerowicz, Elgenius and Phillimore2024; Grønbjerg & Paarlberg, Reference Grønbjerg and Paarlberg2001; Kim, Reference Kim2015). Note that we cannot produce more granular measures of ethnic diversity because of changes in the way this is reported in the census over time.

Our aim is to analyze long-run trends in charity density from 1971 to 2021, with a particular focus on variation in these trends between local authorities as a function of their characteristics. We use growth-curve models in a multilevel modeling framework to estimate the average change in charity density across the census periods, as well as local authority-level differences in this change. These models account for (i) how units of analysis change over time and (ii) between-unit differences in change simultaneously (Singer & Willett, Reference Singer and Willett2003). There have been recent calls for increased applications of these methods in nonprofit and voluntary research (see, e.g., Zhao et al., Reference Zhao, Galaskiewicz and Yoon2022). Growth-curve models have two main factors: an intercept (e.g., baseline level of charity density) and slope (e.g., rate of change in charity density over time; Curran et al., Reference Curran, Obeidat and Losardo2010). They are a specific application of the random-coefficients multilevel model where the coefficient of time (e.g., census year) is allowed to vary randomly between units (e.g., local authorities; Rabe-Hesketh & Skrondal, Reference Rabe-Hesketh and Skrondal2022). The growth-curve model for charity density is specified as follows:

$$ y_{it}={\displaystyle \begin{array}{l}{\beta}_{0}+{\beta}_{1}{period}_{it}+{\beta}_{2}{period}_{it}^{2}+{\beta}_{3}{town}_{it}+{\beta}_{4}{town}_{it}^{2}\\ +\quad {\beta}_{5}{educ}_{it}+{\beta}_{6}{nonwhite}_{it}+{\beta}_{7}{region}_{i}+{\beta}_{8}{rurality}_{i}\\ +\quad {\mu}_{0i}+{\mu}_{1i}{period}_{it}+{\varepsilon}_{it},\end{array}} $$

where y represents the level of charity density in local authority i for census period t, B₀ the mean level of charity density in 1971, B₁ the mean linear change in charity density over time, B₂ the mean curvilinear change in charity density over time, B₃ the mean linear change in charity density for a one-unit increase in material deprivation, B₄ the mean curvilinear change in charity density for a one-unit increase in material deprivation, B₅ the mean difference in charity density for a one-unit change in the percentage of non-White individuals, B₆ the mean difference in charity density for a one-unit change in the percentage of individuals with the highest level of qualification, B₇ the mean differences in charity density for regions outside of London, B₈ the mean differences in charity density for areas that are not classified as “Most Urban,” u₀ the difference between mean level of charity density in 1971 (B₀) and the level for local authority i in 1971 (random intercept for each local authority), u₁ the difference between the mean linear change in charity density over time (B₁) and the linear change for local authority i (random slope for each local authority), and e the difference between observed and predicted charity density in local authority i for census period t.