A typology of social entrepreneuring models continued: Empirical evidence from South Africa

ABSTRACT There have been calls in social entrepreneurship research to move beyond conceptual arguments and to ground definitions and conceptualizations of this field in empirical research, especially pertaining to the developing world. Owing to the socially embedded nature of social enterprises, the context in which social enterprises originate is a key determinant of their modus operandi. In South Africa, a context fraught with social ills, the lack of clarity on the nature and form of social enterprises constrains research and policy formulation. Using a survey methodology, we collected data from a sample of 453 social enterprises on domains of social entrepreneuring identified in previous studies and performed a cluster analysis to identify different types of social entrepreneuring models. The findings point to the existence of two main types of social entrepreneuring models in South Africa: beneficiary-centric entrepreneurial nonprofits and customer-centric social businesses. This paper contributes a more contextual understanding of social entrepreneuring models in South Africa. By showing that social enterprises in South Africa are partly unique to their context, this paper underscores the theoretical and empirical importance of the context in which social enterprises originate and operate when testing the universal validity of social entrepreneuring models.


Introduction
With its dual focus on social mission and economic return, social entrepreneurship is regarded as a model that can strengthen both the economic and social pillars of a country (Littlewood & Holt, 2018;Steinman, 2010;Urban, 2008;Visser, 2011). In developing-country contexts, such as South Africa, social entrepreneurship has a pivotal part to play since government is often unable to meet the vast social needs that exist in the population, and markets do not have the capacity to provide for all those looking for employment and economic security (Littlewood & Holt, 2018;Rivera-Santos et al., 2015). The South African government was initially seen as averse to engaging with social entrepreneurship, regarding social entrepreneurs as "innately risky", and their work as "maverick endeavors" (Urban, 2008, p. 347). More recently, however, social entrepreneurship is receiving higher-level attention in government, not least because there is a compelling argument for the achievement of social impact by financing entrepreneurial activities that address social needs (Jones et al., 2018). The South African government has articulated the need to build inclusivity in the South African economy in order to tackle the nation-wide growing inequality, poverty and unemployment (Barnard, 2019). As a result, the government has recently made concerted efforts to develop a social economy policy (International Labour Organization, 2021).
Given the need and the relatively low uptake of social entrepreneurship in the South African populationthe Global Entrepreneurship Monitor finds that fewer than 2% of adults are involved in social entrepreneurship activity in South Africa, compared to 3.2% globally (Bosma et al., 2016) it stands to reason that the importance of social entrepreneurship would have given rise to extensive research on the matter. However, the literature on social entrepreneurship in South Africa is uneven, though the role of social entrepreneurship in addressing societal development is broadly acknowledged (Jankelowitz & Myres, 2019;Littlewood & Holt, 2018). Currently, the social entrepreneurship debate lacks evidence as to what form social enterprises take, how they operate and the socio-economic benefits they generate. As a result of this, policy development has been slow and legislative barriers and red tape remain (Littlewood & Holt, 2018;Steinman & van Rooij, 2012). At a practical level, social enterprises remain isolated and struggle to access the resources required to realize the benefits that their activities can accomplish in a developing economy (Steinman & van Rooij, 2012).
In general, social enterprise research in the developing-country context is under-developed (Gupta et al., 2020). Recently, scholars declared that "there is much about social entrepreneurship in South Africa that we still don't know" (Littlewood & Holt, 2018, p. 534). Apart from systemic issues related to implementation and research, one of the greatest barriers to developing social entrepreneurship is that there is no reliable database of social enterprises working in the sector, an obstacle which also inhibits the conduct of meaningful research and inhibits the development of policy. The lack of tax registration and poor clarity on how social enterprises are classified in South Africa's Broad-Based Black Economic Empowerment (BBBEE) legislation, which governs the investment by business in economic transformation approaches, also hampers progress (Steinman, 2010). Nevertheless, it would seem that social enterprises continue to start and grow in this context, apparently by adapting their legal status, operations and funding models to suit this less-than-ideal contextual complexity (Littlewood & Holt, 2018;Mair, 2020).
There is also little consensus on the boundaries and characteristics of social enterprises (Rivera-Santos et al., 2015). This is because social enterprises are complex organizations and take different forms in different contexts (Chandra & Kerlin, 2020;Kerlin, 2010). The type of tensions and challenges that face the organization will greatly differ between different environments. This complexity can be seen in numerous studies (e.g. Bacq & Janssen, 2011;Dacin et al., 2010;Kerlin, 2010;Mair, 2020;Zahra et al., 2009) reporting many disparate definitions of social entrepreneurship. In this regard, there have been calls to extend existing theories or even generate new ones by investigating contextspecific phenomena, including social entrepreneurship, on the African continent (Chandra & Kerlin, 2020;George et al., 2016).
As Defourny and Nyssens (2017) observe, decades of definitional debates on social entrepreneurship may point to "the impossibility of a unified definition" (2017, p. 2471), which could be solved with empirical studies following a "bottom-up approach". Such an approach would do away with definitional frameworks superimposed from existing legal traditions or organizational conceptualizations, allowing for context-specific organizational models to emerge (Mair, 2020). Scholars have been advocating for meso-level investigations, which would likely be country-specific (Coskun et al., 2019;Defourny & Nyssens, 2017;Littlewood & Holt, 2018) and appropriate for contextslike South Africalacking an existing legal framework for social enterprises. However, even though such bottom-up approaches are present on an international scale (e.g. Defourny et al., 2021;Mair et al., 2012), they are largely absent for South Africa and Africa at large (Defourny et al., 2021).
In this study, we attempt to address some of the research gaps identified above by analyzing the data from a comprehensive quantitative survey of social enterprises in South Africa (Myres et al., 2018). We endeavor to uncover the nature and characteristics of social enterprises by following the trail of research on typologies of social entrepreneuring models. A sample of 453 social enterprises was drawn from existing social entrepreneurship networks in South Africa. Using a survey methodology, the data was collected on domains of social entrepreneuring identified in previous studies (e.g., Doherty et al., 2014;. These domains include: organizational characteristics, activities, constituencies, financial status, funding mechanisms and growth plans. We performed a cluster analysis to identify different types of social entrepreneuring models, allowing for the social entrepreneuring model types to emerge inductively from the data rather than deductively from a pre-existing theoretical model. The findings point to the existence of two main types of social entrepreneuring models in South Africa, namely: beneficiarycentric entrepreneurial nonprofits and customer-centric social businesses. To our knowledge, this is the first empirical quantitative study focusing specifically on social entrepreneuring models in South Africa. This paper contributes to practice a contextual understanding of social entrepreneuring models in South Africa (Chandra & Kerlin, 2020), based on empirical evidence from a considerable sample of social enterprises and a comprehensive list social entrepreneuring domains. By showing that social enterprises in South Africa are partly unique to their context, and similarly to previous empirical studies in this geographic region (Littlewood & Holt, 2018;Rivera-Santos et al., 2015), this paper considers the operational practices of social enterprises while underscoring the theoretical and empirical importance of the context in which social enterprises originate and operate when testing the universal validity of social entrepreneuring models (Chandra & Kerlin, 2020;Mair, 2020).

Background
Social enterprises are not a new phenomenon in South Africa (Littlewood & Holt, 2018). In 1991, Ashoka established an office in the country. In 2003, the BBBEE Act, which underlies the country's efforts towards transformation and the empowerment of previously disadvantaged individuals (Littlewood & Holt, 2018), was promulgated and then amended in 2013. The African Social Entrepreneurs Network (no longer existing) was created in 2009, followed by the Social Enterprise Academy's South African branch in 2012. Both African networks maintained strong relationships with UK-and US-based organizations like Ashoka, the Schwab Foundation for Social Entrepreneurship and the International Centre for Social Franchising. The Bertha Centre for Social Innovation and Entrepreneurship was launched in 2011 at the University of Cape Town. In 2009 the University of Johannesburg, with support from the International Labour Organization (ILO) and the Belgian government, embarked on a study of 24 social enterprises in South Africa, covering the business models, challenges, types of product/service offered and target market of social enterprises (Littlewood & Holt, 2018). The University of Pretoria's Gordon Institute of Business Science (GIBS) initiated a Social Entrepreneurship Program in 2009, which has run continuously every year since, providing training and support to over 600 social enterprises. In 2015, The Government of Flanders funded a multi-year program of social entrepreneurship development and research, in partnership with GIBS, one of the outputs of which was the study (Jankelowitz & Myres, 2019;Myres et al., 2018) on which this paper is based.
The above-mentioned initiatives and institutions have brought some legitimacy to social enterprises in South Africa, so much so that the South African government is currently drafting a social economy policy (International Labour Organization, 2021). However, since there is no legal framework for social enterprise in South Africa, there is no way to be certain how many of these organizations exist. Some operate as nonprofits, but others as for-profit enterprises; still, others operate with blended legal forms to achieve their blended social and economic missions (Claeyé, 2017), apparently because they offer greater operational flexibility (Mair, 2020).
The above scenario introduces the macro-level institutional framework for social enterprises in South Africa. According to Kerlin's (2012) macro-institutional social enterprise (MISE) framework, a country's macro-level institutionssuch as the welfare state, economic competitiveness, civil society, international aid and culture (Coskun et al., 2019) influence the organizational models social enterprises adopt in a particular context and explain the cross-country variations we observe in social entrepreneuring models, a view also empirically supported by Mair (2020). The MISE framework was developed departing from social origins theory and historical institutionalism. According to the former theory, countries take different routes of third-sector development based on the interplay between state, market and nonprofit provision, which is influenced by historical trajectories (Salamon & Anheier, 1998). In a similar line, historical institutionalism posits that "institutions emerge from and are sustained by features of the broader political and social context" in a specific country (Thelen, 1999, p. 384). Applying these theories to the social enterprise sector, Kerlin (2012) postulates that national contexts exhibit varying social entrepreneuring configurations owing to different histories of culture, governance mechanisms and political-economic structures.
Returning to South Africa's social-enterprise context, Kerlin (2012) classified South Africa's economy as efficiency-driven, in line with the Global Entrepreneurship Monitor (Bosma et al., 2016), meaning an economy characterized by industrialization with state policies aimed at supporting efficiency and product quality, favoring large business over small business and entrepreneurship. This is complemented by South Africa's reportedly mediocre entrepreneurial ecosystem (Bowmaker-Falconer & Herrington, 2020); however, one may argue that South Africa, with its dual economy, exhibits elements of both factor-driven and innovation-driven economies. From the perspective of civil society, Kerlin (2012) positions South Africa as borderline between the traditional and deferred democratization model of civil society. According to the former, South Africa would have a small civil society sector with little government support, a focus on addressing poverty, and reliance on fees, international aid and volunteerism. At the same time, the country also displays elements of the latter model, whereby civil society remains small due to policies still focusing on mainstream economic development. Almost ten years later, however, this scenario is changing, with the South African government and other institutions recognizing the social economy as legitimate.
From the point of view of the MISE framework, the macro-level institutional dimensions of economic development and civil society shape the organizational configurations of social enterprises in a given country (Kerlin, 2012). Based on the above, South Africa would present a sustainable subsistence model of social enterprise with some elements of the autonomous mutualism model. This seems to be the case empirically, with social enterprises in South Africa being small in scale and taking on predominantly the nonprofit organizational form, and with no social enterprise policy and no dedicated social enterprises legal framework in placeat the same time exhibiting some autonomous mutualism model elements, such as the importance of co-operatives in the social economy and a certain degree of variation in social-enterprise activities (Myres et al., 2018).
After looking at the South African macro-institutional context for social entrepreneurship, we turn our attention to meso-level, organizational dimensions.

Management Methods and Approaches
Given that social enterprises often operate in highly complex and severely resource-constrained environments, it is useful to consider how they function at the organizational level. However, this represents an under-researched area. Sassmannshausen and Volkmann (2018) performed a systematic review of the literature and noted that only 20% of all published articles focus on social enterprises from an organizational theory perspective. A more recent review has similarly found that the social enterprise literature has yet to investigate business models and marketing strategies used in SE in any detail (Gupta et al., 2020). Further, it has been argued that a focus on internal organizational features will add to understanding of the phenomenon of social enterprise (Mair, 2020).
Social enterprises are not entirely focused on traditional commercial objectives and outcomes (Mair et al., 2012). Instead, social enterprises prioritize the concerns of the communities they serve or strike a balance with these and the need to manage costs and optimize profits. Success in this invariably requires imitation of the operating logics of forprofit entities in order to capture market attention and to be able to compete. As a result, such hybrids tend to be complex organizations, if not in scale, certainly in activity, requiring distributed agency for success and sustainability (Gupta et al., 2020;Mair et al., 2012). Arguably, navigating these contradictory tensions demands boldness in defining and assuming institutional and cultural identities that evolve as the purpose changes in scope (Mair et al., 2012).
Social enterprises break convention in multiple ways. They often span sectors both in terms of legal organizational conventions (Dorado & Ventresca, 2013;Mair, 2020), as well as technical operating competency requirements. As an extension of their hybrid contextspecific structures, social enterprises have to define an identity that best serves their goals. The leadership must make sense of all involved parts in a bid to define values that motivate and inspire followership by internal and external stakeholders. This may be in the form of a charismatic founder or the recruitment of experienced managerial teams and directors to steer the development of comprehensive long-term plans (Diochon & Anderson, 2009;Waddock & Post, 1991).
The management of social enterprises is complex, in keeping with the base missions that inform the founding and sustainability of these firms. For example, in an empirical study, Tracey and Phillips (2016) critically examined the managerial experiences of Keystone, an organization that set out to support migrant communities in the UK. The case study describes the challenges caused by the negative stigmatization of trying to serve a marginalized group in a way that delivers positive outcomes for all stakeholders. As a result, community interactions often manifest as complex and affected by peculiarities that present unique management challenges. Consequently, managers in social enterprises must have the capacity to learn and pragmatically evolve as stakeholder expectations and/or market conditions change. They must often contend with resentment and resistance from the very communities that they seek to serve (Tracey & Phillips, 2016).
Other management challenges experienced in social enterprises can be broadly grouped into four categories, namely performing, organizing, belonging and learning (Smith et al., 2013). Performance complexities arise from the difficulties associated with the competing goals that must exist to satisfy the needs of diverse stakeholders. Organizing tensions emerge when social enterprises either integrate or separate social and business activities and because of the need to recruit and manage employees with a diverse range of skills and competencies. Establishing a clear organizational identity creates a belonging tension, since different stakeholders hold different expectations in this regard. Finally, social enterprises must balance the need for short-term financial gain with the need to achieve longer-term social impact, thereby creating tensions in learning and competency development (Smith et al., 2013).

Typologies of Social Entrepreneuring Models
Owing to the socially embedded nature of social enterprises (Mair & Martí, 2006), the context in which social enterprises originate is a key determinant of their modus operandi (Rivera-Santos et al., 2015). In light of this, studying a western-born theoretical field such as social entrepreneurship in novel contexts of application such as Africa has the potential to contribute to or challenge existing models (George et al., 2016;Gupta et al., 2020;Zoogah et al., 2015).
In South Africa, there is a lack of clarity on the nature and form of social enterprises, which is one of the causesor, arguably, one of the consequencesof the lack of a legal framework for these organizations (Sengupta et al., 2018). Hence, there is a practical need to know what social enterprises look like in this country in order to inform policies and future research. More specifically, it is also important to know and appreciate the differences between social enterprises operating in the same national context, as social entrepreneuring models not only differ across countries (Kerlin, 2012) but also within countries (Mair, 2020).
We shall now dwell on previous research attempting to develop a typology of social entrepreneuring models. Given the contextual focus of the present study, we shall consider South African or pan-African typologies and cross-country typologies including South African social enterprises. We will consider both conceptual and empirical efforts. Among the former, and moving from broad/cross-country to narrow/national studies, we situate Defourny and Nyssens (2017) theoretical framework encompassing four social-enterprise models: i) the entrepreneurial non-profit model; ii) the social cooperative model; iii) the social business model; and iv) the public-sector social enterprise model. Each of these models represents a transition of a mutual-interest, capital-interest or general-interest organization to more blended-interest logics where hybrid resources are employed. For instance, the entrepreneurial non-profit emerges when general-interest non-profits start complementing public grants and donations with earned-income sources.
Another worthy conceptual effort is Margiono et al.'s (2018) social venture business model configurations. From the perspective of resource dependence theory and public administration theory, they identified three business modelslock-in centered, novelty-centered and efficiency-centeredbased on whether funding and external control are public or private, leading to different configurations of value creation and value capture and to different degrees of autonomy and legitimacy.
Finally, focusing on South Africa, Claeyé (2017) offered a conceptual typology of social enterprises based on the legal forms they may adopt under current national legislation, as described in an earlier section. He distinguished between not-for-profit entities (voluntary associations, trusts and non-profit companies), for-profit entities (private companies, personal liability companies, public companies, close corporations, co-operatives and sole proprietorships) and hybrid structures (a combination of not-for-profit and for-profit configurations). Through adopting a hybrid structure, social enterprises divide their aims, objectives, and activities between two or more legal entities; for instance, a forprofit entity such as a co-operative or a private company with a not-for-profit one such as a voluntary association, trust or non-profit company. While such hybridity increases the complexity of a social enterprise (Claeyé, 2017), it also offers the enterprise the flexibility to raise funds from multiple sources (Mair, 2020).
Empirical studies on typologies of social entrepreneuring models are also available and will be reviewed next. These hold great potential in that they illustrate how social enterprises are actually organized in different contexts, while conceptual frameworks help us to make sense of such organizational configurations. Starting with studies more international in scope, Mair et al. (2012) conducted an empirical study of social entrepreneuring models using a global sample of 200 social entrepreneurial organizations from the Schwab Foundation and Ashoka. Their cluster analysis revealed the existence of four social entrepreneuring models, each leveraging different types of capitalpolitical, human, economic and socialand each having different target issues, constituencies and actions, as well as exhibiting different logics to justify their action. However, this study's sample is not representative of social entrepreneurial organizations (Mair et al., 2012, p. 357) and only includes 24 organizations from Africa. Defourny et al. (2021) tested Defourny and Nyssens (2017) theoretical typology using a sample of 721 social enterprises from 43 countries, including 55 organizations from South Africa and Rwanda. Their hierarchical cluster analysis yielded seven clusters, which the authors could reconcile with three of their four social-enterprise models, namely: the entrepreneurial non-profit, social co-operative and social business model. Still, as the authors mention, there is an uneven distribution across continents in their sample, "with a quasi-absence of Africa" (Defourny et al., 2021, p. 8).
Closer to this present study's geographical context, Littlewood and Holt (2015) surveyed 270 social and environmental enterprises in Southern and Eastern Africa, with Kenya and South Africa being the most represented countries in their sample. The study has merit in that it compares and contrasts social and environmental enterprises in Africa along some key characteristics. For instance, African environmental enterprises appear to operate more similarly to traditional business, with trading income as their main income source, than social enterprises, which rely more on grants, donations and membership fees. Their sources of start-up funding also differ, with environmental enterprises largely being funded from owners' savings and investments and social enterprises relying relatively more on funding from international charities and aid agencies. A downside of this study, however, is that it superimposes the distinction between social and environmental enterprises as opposed to applying more sophisticated methods to allow a typology to emerge from the data.
We situate our study along this stream of empirical investigation of social entrepreneuring models representing meso-level research using a bottom-up analytical approach in search of "real-types that complement ideal-type schemes" (Mair, 2020). With the present study, we seek to obviate for either the lack of methodological rigor in analyzing typologies of social entrepreneuring models in Africa or the under-representation of African social enterprises in international samples. Recognizing that using such a meso-level approach is bound to generate country-specific typologies (Defourny & Nyssens, 2017), we advance the above literature by delving into a specific country context (South Africa) using a quantitative methodology similar to the one employed in cross-country empirical studies.

Sampling Method
The study's population was defined as all social enterprises operating as formal-sector organizations (Myres et al., 2018). The study aimed to achieve a national spread of social enterprises from all nine provinces of South Africa. Respondents were either the enterprise's founder or a manager. The study adopted a purposive quota sampling strategy, in line with similar studies (e.g., Bacq & Eddleston, 2018;Urban & Gaffurini, 2018).
It is generally agreed that there is currently no comprehensive, available sampling frame or database of social enterprises in the country, and therefore the study commenced with the construction of a list of organizations from which the sample could be drawn. This involved the use of a privately held list of civil-society organizations and the conduct of a social media campaign inviting social-enterprise participation in the study. Respondents were also asked for referrals. This procedure resulted in a list of 33,784 enterprises.
To draw the study's sample, each of the organizations included in the ad-hoc sampling frame was contacted telephonically and asked the screening questions based on the sampling criteria. In order to cover both the social and entrepreneurial aspects, the sampling selection criteria included: i) the enterprise prioritizing its social and environmental mission; ii) the entrepreneur identifying his/her organization as a social enterprise; and iii) more than 25% of revenues being from trade or memberships. This sampling procedure and the ensuing data cleaning procedure led to the final 453 final responses obtained in this study.

Survey Questionnaire
The data collection instrument was a survey questionnaire largely consisting of ad-hoc scales developed to suit the context of investigation. It comprised 34 self-reporting statements measuring the social enterprise's organizational characteristics, activities, constituencies, financial status, funding mechanisms, growth plans and future challenges. Owing to the different nature of the social entrepreneurial aspects being measured, the questionnaire was made up of a combination of different levels of measurement, yielding ratio, interval ordinal and nominal data.
The questionnaire was field-tested, and adaptions made to survey questions where appropriate. Responses were collected by administering the survey telephonically (69% of responses), in a face-to-face interview (23%), or through an online link (8%). The data was collected over a four-month period (Myres et al., 2018).

Data Analysis Methods
In a first stage, we analyzed the structure of the data by conducting data reduction techniques in line with the nature of the data. In a second stage, we conducted cluster analysis to identify different types of social entrepreneuring models. In a third and final stage, we performed discriminant analysis, t-test and Pearson's chi-square test of independence to validate the cluster analysis results.
As for analyzing the structure of the data (stage 1), and in order to prepare the data for cluster analysis, it was needed to first apply data reduction techniques. For interval data (items measured on a 5-point Likert scale) we conducted exploratory factor analysis. After conducting reliability analyses on the resulting factors, we only kept the factors with an internal consistency Cronbach α score higher than 0.6 (ranging from 0.629-0.853), which is an acceptable reliability score threshold considering that the factors were composed of 3-7 items each (Field, 2013). For nominal data, we followed an optimal scaling procedure and applied categorical principal component analysis in order to obtain composite scores. This procedure resulted in count variables, representing the sum of the presence of a variable included in the dimensions as identified in categorical principal component analysis.
Next, we conducted cluster analysis (stage 2). Cluster analysis is an exploratory analysis technique that attempts to identify structures within the data by grouping a set of objects or individuals in homogenous clusters. Given the categorical and continuous nature of the data, we conducted two-step cluster analysis as the appropriate clustering method (Norušis, 2012). As the name suggests, two-step cluster analysis entails two steps: in the first step, cases are assigned to pre-clusters according to a modified hierarchical agglomerative clustering procedure that combines the objects sequentially to form homogenous clusters following a so-called cluster feature tree (Mooi & Sarstedt, 2010). In the second step, the pre-clusters are clustered using the hierarchical clustering algorithm (Norušis, 2012). Finally, the procedure suggests how many clusters to retain from the data based on Bayes' Information Criterion, which avoids the arbitrariness of traditional clustering techniques (Mooi & Sarstedt, 2010;Norušis, 2012).
Finally, to validate the cluster analysis results (stage 3), we ran discriminant analysis on the clustering solution results, as well as t-test and Pearson's chi-square test of independence to verify the existence of statistically significant relationships between the cluster membership variable and the variables included in the cluster analysis. To further validate the cluster analysis results, we also performed t-test and Pearson's chi-square test of independence for the variables excluded from the cluster analysis.

Results
The first stage in data analysis consisted in applying data reduction procedures (exploratory factor analysis and optimal scaling), as outlined in the previous section. This resulted in 47 attributes that could be included in the cluster analysis.
The second stage in the data analysis procedure was represented by two-step cluster analysis. This entailed first establishing which items to include in the analysis. When making this choice, we kept in mind that we needed the clusters to be representative of social enterprises in the key domains that were measured, i.e., organizational characteristics (8 items), activities (15), constituencies (9), financial status (4), funding mechanisms (5), growth plans (3) and future challenges (3). We performed a first evaluation of the total set of items and arrived at list of 27 items that were deemed most representative of each social entrepreneuring domain investigated. Next, we allowed the cluster analysis to determine how many items should be included in the final clustering solution, bearing in mind that cluster analysis provides indicators of which clustering solution is the most reliable based on which attributes are included in the cluster analysis as well as the number of clusters resulting from the clustering procedure. Hence, different cluster analysis iterations were performed with the number of attributes included ranging from 10 to 27, in order to determine the clustering solution with the best value of cohesion and separation. This iterative procedure resulted in a final clustering solution with 24 segmentation variables. The variables were then screened again to determine whether they were representative of the social entrepreneuring domains investigated. It was deemed that, out of 47 items available after the factor analysis and optimal scaling procedures, the 24 attributes included in the final clustering solution were representative of the key social entrepreneuring domains investigated. Table 1 presents the variables included in the final clustering solution. The final chosen clustering solution with 24 separation variables had a silhouette measure of cohesion and separation 1 of 0.1, which is an indication that the final clustering solution had a poor goodness-of-fit quality. However, the two-step clustering procedure could classify all 453 respondents into two clusters. Table 2 presents the final clustering solutions results. (2) Less than 3 months; (3) More than 3 months, less than 18 months; (4) More than 18 months, less than 3 years; (5) More than 3 years, less than 5 years; (6) More than 5 years Scope of operations Ordinal (1) Local community based; (2) Across a single province; (3) Across multiple provinces; (4) National, across the whole country; (5) Regional, across several countries nearby; (6) Global, across many countries Activities Industry Nominal (1) Accommodation and food service activities; (2) Administrative and support service; (3) Agriculture, forestry and fishing; (4) Arts, entertainment and recreation; (5) Construction; (6) Education and training;  (1) Access to public services such as transport, energy, water and sanitation; (2) Access to support and advisory services; (3) Crime in the area in which you operate *Subjected to optimal scaling and recoded as a count variable, whose score represents the sum of the presence of a variable included in the dimensions as identified in optimal scaling. **Subjected to factor analysis, where the factor's score was obtained by determining the sum of the rating values for each of the items included in a factor divided by the number of items, thereby obtaining the average value for each respondent. In the third data-analysis stage, we proceeded with different tests to validate the final clustering solution. We performed a discriminant analysis on all the 24 variables resulting from the clustering solution, including the categorical variables (see Table 1), which were recoded into dummy variables. The discriminant analysis results indicated a one-factor solution explaining 100% of the variance in the data, with Wilk's Lambda Λ = 0.218, χ 2 (64) = 638.357, p = .000, and classifying 96.0% of the items correctly. These results were confirmed by the results of an independent-group t-test we ran on all the continuous variables, which revealed statistically significant differences (p = 0.000) for all the continuous variables across the clustering variable, except for number of volunteers (t(451) = 1.467, p = 0.143). To further validate these results, we subjected the categorical variables included in the cluster analysis to Pearson's chi-square tests of independence. Also in this instance, the clustering solution was confirmed for all the categorical variables, except for profit or surplus in previous year (χ 2 (1) = 1.621, p = 0.203).
Finally, we further validated the final clustering solution by testing for statistically significant differences between the clusters in terms of 14 variables excluded from defining clusters. For the seven continuous variables considered, we performed independentgroup t-tests, which revealed that there were no statistically significant differences between the means of the variables external to the defining clusters across the two clustersexcept for core-competencies-based innovation, for which t(451) = -3.079, p = 0.002. For the seven categorical variables excluded from the defining clusters, instead, we performed Pearson chi-square tests. Once again, these revealed no statistically significant relationship with the defining clusters, except for province (χ 2 (8) = 31.227, p = 0.000), rate of growth (χ 2 (4) = 9.790, p = 0.044), and most of the items measuring income sources. Based on the cluster analysis results and all the above validation test results, we concluded that the clustering solution was, overall, valid. The next two sections describe the two clusters more in detail.

Social Entrepreneuring Model 1: Beneficiary-centric Entrepreneurial Nonprofits
The first cluster represents 52.1% (n = 236) of the sample and largely features organizations taking the form of non-profit companies and community organizations. This group is dominated by organizations operating for more than five years and whose operations are largely local-community-based. They operate mainly in education and training but also account for the large majority of enterprises in accommodation and food services, as well as human health and social work. Their activities are more socially oriented, in the sense that to a large extent they do not focus on selling goods or services to customers, but rather center their activities around their beneficiarieswhether it is for providing good/services to them, developing skills, creating jobs, improving health and wellbeing, etc. They also engage more formally with their beneficiaries, which are mainly local communities and particularly disadvantaged groups such as women, youth and the disabled. These organizations self-reported having achieved the following in the previous 12 months: i) beneficiary market growth (e.g., having expanded to include more beneficiaries), ii) organic growth (e.g., having increased market share relative to competitors), and iii) funding goals (e.g., having attracted funding from new donors). These organizations' innovation efforts are centered around their product/service offering to beneficiaries and customers. They also have more deliberate practices in terms of the monitoring and evaluation of their social impact: they have established internal monitoring and learning systems, produce integrated reports of social and environmental impact, and monitor their social impact on a weekly, monthly or at least quarterly basis. In terms of their funding sources, even though the majority (31.8%) of these organizations do not receive donations, the rest have an even spread of proportion of grants and donation in their funding model and, in fact, rely more heavily on non-commercial funding. Finally, compared to the other cluster, they report a more significant presence of challenges such as access to infrastructure, access to support and advisory services, and crime.

Social Entrepreneuring Model 2: Customer-centric Social Businesses
The second cluster represents 47.9% (n = 217) of the sample and groups together organizations mainly structured as social enterprises (30.9% of the cluster), businesses and nonprofit companies. The majority of these enterprises have been operating for more than five years (36.9%), but they are quite spread in terms of this variable and are also fairly young (46.1% operating for less than three years). Likewise, while their operations are mainly local-community-based, they also have wider (provincial, national and even global) reach. In terms of the industries in which they operate, they mainly operate in the education and training sector but also dominate in the arts, entertainment and recreation, as well as traditional commercial sectors such as agriculture, information and communication technology, manufacturing, and wholesale and retail trade. In their activities, these organizations are more customer-centered than socially oriented, as they score higher on all types of customers served, whether disadvantaged/poor consumers, NGOs or even large corporates. Likewise, these organizations have a less formalized and less structured interaction with their beneficiaries. They place comparatively less emphasis on increasing their services to, or growing their base of, existing beneficiaries, who are mainly defined as particular locations, organizations or employees. Perhaps surprisingly, the organizations in this cluster are less focused on growing by improving product/service quality, increasing market share or entering into new partnerships. When it comes to measuring social impact, these organizations also have established internal monitoring and learning systems, but their monitoring and evaluation practices vary greatly and include organizations that expend "no effort and resources" and "measure activity outputs only". Moreover, while the majority (65.9%) of these organizations declared having evaluated their impact on beneficiaries in the previous quarter, most of the remaining organizations (17.1%) declared never evaluating their impact. Interestingly, given their pronounced customer-centric and commercial focus, organizations in this cluster employ roughly eleven volunteers on average, i.e., six more than those in the other cluster. From a financial and funding perspective, they stand out in their predominant under-reliance on grants or donations and their reliance on commercial funding.

Discussion
This paper sought to uncover the typology of social entrepreneuring models in South Africa using methods that allowed such typology to surface from the data rather than applying a theoretically derived framework. This approach resulted in two types: beneficiary-centric entrepreneurial nonprofits and customer-centric social businesses. The labels applied to these two social-enterprise types were devised considering a constellation of characteristicsrather than a dominant featureand the comparable international literature on typologies of social entrepreneuring models. We discuss the two social entrepreneuring models and their characteristics in the following paragraphs. In the process of coming up with the most representative and comprehensive names for the two clusters, we considered the following factors. In the first place, we felt the need to capture the fact that the most powerful clustering variable was the organizational type of the social enterprise (see Table 2), a variable that also allows the drawing of comparisons with international typologies. Secondly, it was evident to us that the first cluster scored higher on a number of beneficiary-focused dimensions, while the second cluster scored higher on all customer-focused statements (see the Results section and Table 2). This beneficiary/customer focus seems to capture and run through a constellation of dimensions, from governance structures and activities to monitoring-and-evaluation practices and funding sources.
The first social entrepreneuring model, which we termed beneficiary-centric entrepreneurial nonprofits, seems to fall under what we typically understand as a not-for-profit, with a clear reliance on grants and donor funding, a concerted focus on prioritizing the interests of their beneficiaries, and a deliberate effort to monitor and evaluate their social impact. The latter is a significant finding, given the previously reported lack of social impact measurement practices in South Africa (Urban, 2015). These social enterprises resemble Littlewood and Holt's (2015) "social enterprises" in Africa, with a heavy reliance on donor and grant funding. Internationally, they seem to be close to Defourny et al.'s (2021) entrepreneurial non-profit model, made up mainly of not-for-profit organizations; with a mission to improve access to education, healthcare services, equality and employment (also through work-integration models); relying mainly on subsidies and donations as income sources; and having democratic governance models involving beneficiaries. Because of the centrality of beneficiaries and the resemblance with Defourny et al.'s (2021) entrepreneurial non-profit model, we named this social entrepreneuring model "beneficiary-centric entrepreneurial nonprofits".
The second type, which we termed customer-centric social businesses, may seem less intuitive and offers a number of valuable insights into the nature and practices of social entrepreneurship in South Africa. What stands out is that this social entrepreneuring model is more hybrid in nature: the majority of these organizations takes the organizational form of either a social enterprise or a business, but there is also a non-negligible number of non-profit companies in this cluster. Additionally, while self-identifying as social enterprises, these organizations are more customer-oriented and also operate in mainstream economic sectors such as agriculture, ICT, manufacturing and retail. Drawing comparisons with international typologies, this social entrepreneuring model is akin to Defourny et al.'s (2021) "social-business model" and Erpf et al.'s (2019) "social service providers", represented by for-profit legal forms with a significant reliance on trading income and commercial sources of funding, and less involvement of beneficiaries in governance structures. Because of their customer orientation on a number of dimensions, it is likely that these social enterprises exact some fees from their beneficiaries, and hence also view them as customers, which is typical of an integrated economic model of social enterprise (Saebi et al., 2019) and characteristic of countries with a similar civil society model and stage of economic development (Kerlin, 2012). Owing to its clear attributes of organizational hybridity, market orientation and customer focus, we named this social entrepreneuring model "customer-centric social businesses".
As noticed earlier, customer-centric social businesses are younger than their non-profit counterparts, which may raise the question of whether, in South Africa, there is a transition from the customer-centric social-business to the beneficiary-centric entrepreneurialnonprofit type over time. These organizations' more commercial nature is also in line with the wider scope of their operations, which may even reach global markets. Moreover, given their more commercial or market orientation, customer-centric social businesses do not put much effort into measuring social impact. We may postulate that this is not so much because they do not have the skills and systems in place to measure impact, but more because they focus more on achieving and measuring their economic goals using traditional economic measures to assess their progress and impact. Hence, in light of the above discussion, the "entrepreneurship" aspect (broadly understood as the economically sustainable aspect of a social entrepreneurial business model) of social entrepreneurship is more prevalent in this type of social entrepreneuring model. This observation offers more granularity to Littlewood and Holt's (2015) description of social enterprises in Africa being more akin to non-profits: it appears that a portion of social enterprises in South Africa does indeed follow more traditional business logics, similar to what was observed concerning environmental enterprises in their study.
A somewhat unintuitive finding is the presence of a greater number of volunteers in customer-centric social businesses, whereas one would expect the more non-profit oriented organizations to rely more on volunteers. We could ascribe this finding to the supposition that customer-centric social businesses, as classified in this study, are larger organizations than their beneficiary-centric entrepreneurial-nonprofit counterparts; hence, they may employ more volunteers in absolute terms given the larger scale of their operations, but the volunteer-employee ratio may be lower in these organizations than in beneficiary-centric entrepreneurial nonprofits. However, the post-hoc tests we performed (t-test and Pearson's chi-square test of independence) revealed no statistically significant differences across the two types in terms of number of employees and annual income; hence, our supposition needs to be verified in future research on this aspect.

Conclusion
Notwithstanding advances in the conceptual understanding of the phenomenon of social entrepreneurship (Gupta et al., 2020;Lortie & Cox, 2018), there have been calls to move beyond conceptual arguments and to ground definitions and conceptualizations of this field in empirical research, especially using quantitative methods (Gupta et al., 2020), pertaining to the developing world (Sengupta et al., 2018).
Owing to the socially embedded nature of social enterprises (Mair & Martí, 2006), the context in which social enterprises originate is a key determinant of their modus operandi (Rivera-Santos et al., 2015). In light of this, studying a western-born theoretical field such as social entrepreneurship in novel contexts of application such as Africa has the potential to contribute to or challenge existing models (Barnard, 2020;George et al., 2016;Zoogah et al., 2015).
In South Africa, there is a lack of clarity on the nature and form of social enterprises and concomitant lack of a legal framework for these organizations (Sengupta et al., 2018), which complicates policy-making efforts. This paper represents perhaps the first empirical investigation of the typology of social entrepreneuring models focusing exclusively on the sub-Saharan African context -South Africa in particularfollowing a bottom-up approach as advocated by Defourny and Nyssens (2017) and incorporating a description of the internal operating practices of social enterprises (Mair, 2020) to enhance understanding. Its value also lies in the methodology used, which allowed for the typology to emerge from the data rather than working deductively from theoretically preconstructed frameworks. Moreover, the present paper offers a holistic picture of social entrepreneuring models, given the breadth of social entrepreneuring domains investigated, whilst moving beyond presenting results in an aggregated fashion as in previous studies (e.g., . This paper adds to research some evidence on social entrepreneuring models from South Africa, following the earlier, more global, endeavors by Mair et al. (2012) and Defourny et al. (2021) to uncover ideal social-enterprise types empirically. Overall, the findings of this study point to two unique types of social entrepreneuring models in South Africa: beneficiary-centric entrepreneurial nonprofits and customer-centric social businesses. While reconcilable with other international and pan-African typologies to an extent, this South African typology of social entrepreneuring models is rather contextspecific and not completely commensurable with previously known social-enterprise types. Hence, this study has contributed a more granular and context-specific, mesolevel, as well as data-driven, analysis of how social entrepreneurship manifests in a concrete African context.
From a practical point of view, this paper has special policy-making relevance for South Africa (Barnard, 2019); as the government is working toward a social economy policy, the enhanced understanding of how social entrepreneurial activities are modeled in the country is of paramount importance for formulating and enacting an enabling policy for South Africa's burgeoning social economy. Given the absence of a dedicated legal framework in South Africa, the fact that the most powerful clustering variable was the organizational type respondents identified their organization with is noteworthy. The results of this study suggest to policymakers that non-profit companies and community organizations, on the one hand, and social enterprises and businesses, on the other, have key characteristics in common. Those engaged in the training, support and funding of social enterprises in the South African context will similarly be empowered by this data.

Limitations and Future Research
Even though the study's sampling frame included social enterprises from all nine provinces of South Africa, most organizations in our sample were located in the Gauteng province, followed by the Western Cape province. In order to ensure a more representative sample of the population of social enterprises, we recommend using stratified random sampling in future research with the same population. One needs to bear in mind, however, that there is no reliable database of social enterprises in South Africa, and herein lies the difficulty of using probability sampling methods in this context. We conclude with a few possible avenues for future research. An interesting avenue that surfaced in our discussion is whether in (South) Africa there is a transition over time from being a customer-centric social business to being a beneficiary-centric Kerrin Myres is Associate Professor at the Gordon Institute of Business Science (GIBS), University of Pretoria. She teaches and researches entrepreneurship, social entrepreneurship and qualitative methods, and supervises at Masters' and Doctoral level.