1 Introduction

Over the past decades, social entrepreneurship (SE) has attracted significant attention from scholars and practitioners due to its great potential to solve societal problems using market-based solutions (Mair & Marti, 2006; Saebi et al., 2019). It seeks to create social values while generating profits (Austin et al., 2006; Battilana & Lee, 2014; Mair & Marti, 2006). This focus on potentially competing dual objectives reflects the hybrid nature of SE (Battilana & Lee, 2014; Saebi et al., 2019), making it neither fully understood nor taken for granted up until now (Dart, 2004; Nicholls, 2010; Weidner et al., 2019). The legitimacy perspective provides a fundamental framework for understanding the emergence and development of SE, which primarily focuses on institutional pressure and conformity to stakeholder expectations rather than market efficiency and effectiveness (Bruton et al., 2010; Dart, 2004; Huybrechts & Nicholls, 2013). Three dimensions of legitimacy are well documented: pragmatic, moral, and cognitive (Suchman, 1995). To date, scholars have primarily emphasized the role of pragmatic and moral legitimacy in understanding SE (Bruder, 2021; Dart, 2004) and individuals’ engagement in SE (Au et al., 2023; Miller et al., 2012; Zheng et al., 2022), the cognitive legitimacy of SE remains relatively unexplored.

The limited attention that SE scholars have paid so far to cognitive legitimacy is striking. We aim to explore SE’s cognitive legitimacy for two reasons. First, SE researchers assume that it is theoretically impossible to explore the cognitive legitimacy of SE at its initial stage (Dart, 2004). However, organizational ecologists argue that new organizational forms can obtain cognitive legitimacy from similar categories as they begin to grow (Kuilman & Li, 2009; Lewis et al., 2021; Xu et al., 2014). Exploring the sources of SE’s cognitive legitimacy thus challenges previous studies and enriches current SE literature. Second, cognitive legitimacy—the absence of questions or challenges regarding a new entity (Suchman, 1995; Tost, 2011)—is recognized as “the most subtle and the most powerful source of legitimacy” (Suchman, 1995). When new entities gain cognitive legitimacy, individuals will support them regardless of their moral and pragmatic legitimacy (Tost, 2011). New ventures and organizations thus often seek to obtain cognitive legitimacy to make them understandable to consumers (Shepherd & Zacharakis, 2003) and investors (Maier et al., 2023), increasing their survival rate (Rao, 1994) and enhancing stakeholders’ loyalty to the organization (Tyler, 2006; Tyler & Blader, 2005). Social enterprisesFootnote 1 (SEs) often grapple with a series of challenges, including acquiring financial resources, assessing organizational performance, and navigating a non-munificent institutional environment, which primarily stems from its inherent struggle for legitimacy (Bhatt et al., 2019; Gupta et al., 2020). Building a deeper understanding of how SE obtains its cognitive legitimacy is critical for SE to overcome these challenges. Furthermore, scholars have identified that cognitive legitimacy matters in developing new organizational forms, categories, or practices (Kuilman & Li, 2009; Lewis et al., 2021; Xu et al., 2014; Zhao et al., 2018). Focusing on SE’s cognitive legitimacy thereby can enhance our understanding of individuals’ SE engagement and enrich our knowledge about its emergence.

This paper aims to address a lack of understanding of SE’s cognitive legitimacy. We ask: What are the major sources of SE’s cognitive legitimacy, and how do they affect individuals’ decision to engage in SE activity? Building on legitimacy theory and organizational ecology perspectives (Bitektine & Haack, 2015; Kuilman & Li, 2009), we argue that SE can obtain cognitive legitimacy from “legitimacy spillovers” and the perceived density of SEs. On the one hand, organizational ecology theorists suggest that a great recognition of one organizational category increases the legitimacy of similar organizational forms (i.e., legitimacy spillover) (Kuilman & Li, 2009; Li et al., 2007; Suddaby et al., 2017). Especially at the beginning stage of a new organizational form, when its population starts to grow, initial cognitive legitimation for the new organizational form comes from an existing organizational form with well-established legitimacy (Xu et al., 2014). As an exemplar type of hybrid organization (Doherty et al., 2014), SEs share similarities with two well-established organizational categories, “business” and “nonprofit organization.” Thus, evaluators may use the legitimacy of both business and nonprofit categories to interpret SE based on their shared characteristics. On the other hand, as the new organization form develops, it can obtain cognitive legitimacy from its density (Bogaert et al., 2016; Dobrev & Gotsopoulos, 2010; Hannan & Carroll, 1992). Given SE has undergone decades of legitimation in many countries, we expect it can also obtain cognitive legitimacy from its own prevalence as perceived by the inhabitants (Husted et al., 2016; Lewis et al., 2021). However, these two sources of legitimacy are not fully complementary to each other. As both sources of cognitive legitimacy provide an understanding of SE, extra sources of legitimacy are not needed once the practice of SE is widely understood (Taeuscher et al., 2021). The spillover effects are more important in obtaining legitimacy when the organizational form is new (Xu et al., 2014; Zhao et al., 2018). Therefore, we further predict legitimacy spillover effects to be weaker if individuals perceive a higher density of SE activity in their environment.

Empirically, we use the data from the special survey on SE from the Global Entrepreneurship Monitor (GEM) conducted in 2015. We apply multilevel models to test how a nation’s density of new businesses and nonprofit organizations can influence individuals’ likelihood of starting a social venture. The empirical results confirm our hypotheses: national new business and nonprofit organization density, as well as individuals’ perceived SEs density, are positively associated with individual engagement in SE; furthermore, the influence of new business and nonprofit organization density on individuals’ engagement in SE (i.e., legitimacy spillover) is weaker if the perceived SEs density is high. Further analyses suggest that the results are robust.

This paper contributes to SE, organization ecology, and hybrid organization research. First, it improves the understanding of the cognitive legitimacy of SE, obtained through legitimacy spillover from business and nonprofit categories, as well as derived from SE’s prevalence. It shifts previous attention from SE’s pragmatic and moral legitimacy to its cognitive legitimacy, uncovering the significant role of cognitive legitimacy in predicting individual SE engagement. It also provides empirical evidence of how individuals evaluate an entity’s cognitive legitimacy through either evaluative or passive modes and how passive mode dominates the evaluation process (Bitektine & Haack, 2015; Tost, 2011). Furthermore, it contributes to organizational ecology literature by leveraging a multilevel legitimacy perspective (Bitektine & Haack, 2015), thereby explaining how interpopulation (e.g., legitimacy spillover) and intrapopulation processes (i.e., perceived SE density) affect individuals’ action through a cross-level mechanism. Finally, departing from prior predominant arguments that hybridity often confuses audiences (Pache & Santos, 2013; Zuckerman, 1999), our findings suggest that hybrid organizations can receive legitimacy benefits from categories that they straddle based on the above legitimacy and organizational ecology perspectives.

The remainder of this paper is organized as follows. We first introduce relevant concepts and propose hypotheses. Next, we address our data and the methods used in our analysis. We then report the results of our regression analysis and discuss the results of several robustness checks. Finally, we summarize our main findings and discuss implications, limitations, and suggestions for future research.

2 Theoretical background

2.1 The legitimacy of social entrepreneurship

While most societies’ attention to SE has increased, actors in this field still perceive a lack of recognition (Bhatt et al., 2019; Bruder, 2021), especially regarding its pursuit of social and financial objectives simultaneously (Doherty et al., 2014; Pache & Santos, 2013). Previous research has utilized the legitimacy perspective to comprehend the emergence and evolution of SE (Dart, 2004; Ewald Kibler et al., 2018; Miller et al., 2012). The legitimacy view fits this purpose particularly because it emphasizes institutional pressure and conformity to stakeholder expectations instead of market efficiency and effectiveness (Bruton et al., 2010; Dart, 2004). Suchman (1995: 574) defines legitimacy as “a generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definition.” He classifies legitimacy into three dimensions: pragmatic legitimacy based on audiences’ self-interest calculation, moral legitimacy rooted in the audience’s socially constructed value system, and cognitive legitimacy as a reflection of an organization’s comprehensibility or taken-for-granted assumptions.

Academic studies have so far mainly attributed the emergence of SE to its pragmatic and moral legitimacy (Au et al., 2023; Bruder, 2021; Dart, 2004). For instance, researchers argue that SE has gained some pragmatic and moral legitimacy because of the practical value of its outcomes and the dominance of pro-business ideology in the broader social environment, thus contributing to its emergence (Dart, 2004). Furthermore, Miller et al. (2012) propose that both pragmatic and moral legitimacy affect compassion-triggered SE engagement. More recent research continues to adopt a prosocial or moral standpoint when addressing the emergence of SE (Bruder, 2021; Zheng et al., 2022). By contrast, SE’s cognitive legitimacy has received little attention. This study aims to contribute to previous studies by exploring the cognitive legitimacy of SE for two main reasons.

First, exploring SE’s cognitive legitimacy challenges previous arguments in SE literature that it is inappropriate and unnecessary to consider the cognitive legitimacy of SE during its early development (Dart, 2004; Miller et al., 2012). Dart (2004: 421) argues that the social-enterprise form (at that time) was unprecedented and put forward that “at this stage, it is likely theoretically excessive and unwarranted” to explore SE’s cognitive legitimacy. Nicholls (2010) also discusses SE as a pre-paradigmatic field with limited knowledge about its legitimate methods, usefulness, and problems. More recent studies also argue that SE is neither fully understood nor taken for granted as an emerging organizational entity for stakeholders (Chliova et al., 2020; Saebi et al., 2019; Weidner et al., 2019). However, cognitive legitimacy can be considered to be a continuous rather than a binary variable (Fisher et al., 2016; Gardberg & Fombrun, 2006; Tost, 2011); that is, it reflects the degree to which an entity is understandable. Therefore, even if SE is a relatively new phenomenon, it has been around long enough to gain some extent of cognitive legitimacy. Organizational ecologists also posit that organizations can obtain cognitive legitimacy even at their emerging stage (Kuilman & Li, 2009; Lewis et al., 2021; Xu et al., 2014; Zhao et al., 2018). They argue that new entities can gain cognitive legitimacy through both interpopulation and intrapopulation processes. As a new organizational form begins to grow, it obtains initial legitimation from existing organizational forms with well-established legitimacy, especially when there are shared similarities or identity overlaps with these established forms (Audia et al., 2006; Ruef, 2000). The new organizational form further garners increased acceptance and accumulates more cognitive legitimacy from its density as it becomes more popular and grows in size (Bogaert et al., 2016). Thus, even though the SE field has not reached maturity, it has gained a certain degree of cognitive legitimacy (taken for grandness) from multiple audiences due to its increasing popularity over the past decades (Miller et al., 2012; Short et al., 2009). Consequently, we propose that it is appropriate, and perhaps even necessary, to explore the cognitive legitimacy of SE.

Second, cognitive legitimacy, the widespread acceptance of the organization as necessary or inevitable, is considered to be the most powerful source of legitimacy (Suchman, 1995; Tost, 2011). Tost (2011) argues that if an entity or organizational form reaches a certain level of cognitive legitimacy, individuals will support it regardless of its moral and pragmatic legitimacy. Thus, ventures that adopt new entities often need to obtain cognitive legitimacy to attract consumers or investors (Maier et al., 2023; Shepherd & Zacharakis, 2003). Cognitive legitimacy also helps such ventures to overcome liabilities of newness, increase their chance of survival, and acquire resources (Suddaby et al., 2017; Überbacher, 2014). Organizational ecology literature further finds that the cognitive legitimacy of a category contributes to the survival and growth of the organizations that belong to the category (Hannan & Carroll, 1992; Kuilman & Li, 2009; Xu et al., 2014). Thus, a deeper understanding of the cognitive legitimacy of SE may not only support SEs’ survival but also provide insights into developing the whole category of SE. Overall, we aim to contribute to a deeper understanding of where the cognitive legitimacy of SE comes from in order to better explain and perhaps influence individuals’ choices to participate in SE activity (Miller et al., 2012; Townsend & Hart, 2008).

2.2 Hybrid organizations and social entrepreneurship

Hybrid organizations, known to combine aspects of multiple organizational forms, are typically considered to be lacking legitimacy due to their deviations from existing well-established categories that convey a coherent social recognition for a group of organizations (Deephouse, 1996; Wry et al., 2011). The mixed identities, organization forms, and institutional logics in hybrid organizations often confuse what “type” they belong to, which further causes difficulties for them in obtaining and maintaining legitimacy (Greenwood et al., 2011; Kostova & Zaheer, 1999). Social enterprise is often recognized as an exemplary type of hybrid organization that combines elements from both commercial business and nonprofit organizations (Battilana & Lee, 2014; Chliova et al., 2020; Doherty et al., 2014; Pache & Santos, 2013). Hybridity often makes social enterprises deviate from a widely acknowledged type (i.e., commercial business or nonprofit organization), leading to a lack of legitimacy. However, hybridity also exemplifies the shared similarities and identity overlaps between social enterprises and commercial businesses or nonprofit organizations. Specifically, SE is similar to commercial entrepreneurship regarding resources, context, and opportunity (Austin et al., 2006). Both commercial and social entrepreneurs wish to obtain future returns, are embedded in similar external contexts (including tax, regulatory, sociopolitical environment, and macroeconomy), and require financial and human resources. Besides, SE emerges from nonprofit sectors and shares nonprofit organizations’ social and civic orientation (Battilana & Lee, 2014; Dart, 2004; Haugh, 2007).

Considering the shared similarities, research suggests that audiences may at some point tolerate the disruption caused by category spanning (Durand & Paolella, 2013) as categorical boundaries become blurred (Battilana & Lee, 2014) and new categories emerge due to the category-spanning (Durand & Khaire, 2017). This line of reasoning therefore proposes a somewhat different perspective on the link between hybridity and legitimacy in the context of SE that is only limitedly validated via (quantitative) empirical analysis.

3 Hypotheses development

3.1 Legitimacy spillovers and SE engagement

Organizational ecology literature posits that the cognitive legitimacy of new organizational forms can be obtained through the interpopulation process by receiving “legitimacy spillovers” from an existing organizational form with well-established legitimacy (Kuilman & Li, 2009; Xu et al., 2014). In other words, legitimacy spillovers occur when greater social recognition for one established category results in a higher social recognition for related organizations as well (Kostova & Zaheer, 1999; Kuilman & Li, 2009; Li et al., 2007). Such spillovers are usually based on the similarities between the new and well-established organizational forms (Haack et al., 2014; Suddaby et al., 2017). For instance, Kostova and Zaheer (1999) propose that the legitimacy of subunits of a multinational organization (MNE) may be evaluated by the MNE as a whole since they belong to the same cognitive category. Research has also shown that legitimacy spillover can occur in the context of foreign banks (Kuilman & Li, 2009) and financial cooperatives (Dobrev et al., 2006).

Based on the legitimacy spillover perspective, legitimacy can transfer from the well-established organizational form to the new form as long as there is some identity overlap between them (McKendrick et al., 2003; Xu et al., 2014). Thus, legitimacy spillover may also occur for hybrid organizations as they can build parallels or analogies using existing categories they straddle, thereby increasing audiences’ understanding of what they are (Alexy & George, 2013; Martens et al., 2007). Empirical research has confirmed the existence of legitimacy spillovers within certain hybrid organizations or practices (Peng, 2003; Xu et al., 2014). For instance, science-technology-hybrid start-ups are positively evaluated by venture capital (Wry et al., 2014). China’s collectively owned enterprises can gain legitimacy from state-owned enterprises as they share some key identity codes (Peng, 2003; Xu et al., 2014). Drawing on the above perspective, we expect SE may gain legitimacy from spillover effects from both commercial business and nonprofit organizational categories as it combines elements from both categories.

However, prior research mainly explored legitimacy spillover effects at the macro level (i.e., how the density of an organization is positively associated with that of another) (Kuilman & Li, 2009). Researchers have attributed the macro-to-macro relationship to several across-level causal mechanisms (Coleman, 1994; Hedstrom & Swedberg, 1998). As Bitektine and Haack (2015) argue, legitimacy is a cross-level process that includes macro-level ‘collective’ legitimacy judgment (i.e., validity), which influences micro-level perceptions and judgment of social acceptability (i.e., property). The macro-level validity can thus affect how individuals evaluate observed organizational behaviors and properties, which further affects individuals’ behavior (Bitektine & Haack, 2015). Related to this cross-level mechanism, Scott (1995) and Suchman (1995) argue the comprehensibility of a category originates from the availability of cultural models that provide plausible explanations. When an organizational form is new to the evaluator, s/he needs conceptual exemplars to interpret it (Zhao et al., 2018). Commercial businesses and nonprofit organizations, as two well-established categories, thus provide parallels and/or syllogisms to individuals to help them understand SE through shared identities: the legitimacy of the business category assists individuals in understanding SEs’ business-related activities, while the legitimacy of the nonprofit category improves individuals’ understanding of SEs’ social mission. Overall, the commercial business and nonprofit categories can thus make SE predictable, inviting, and meaningful (Bitektine & Haack, 2015; Suchman, 1995). When individuals perceive SE to be understandable (i.e., cognitively legitimate) through the spillover effect of legitimacy from both business and nonprofit organizations, they are more likely to engage in SE creation.

In line with previous studies (Lewis et al., 2021; Suddaby et al., 2017), we proxy the legitimacy of commercial business and nonprofit categories by their population density. A higher prevalence of organizations indicates a higher level of social recognition and acceptance level of these organizations; therefore, more cognitive legitimacy is accumulated for them (Bogaert et al., 2016; Chung & Cheng, 2019). We thus argue that the more legitimacy both commercial businesses and nonprofit organizations have, the more easily an individual can be expected to understand SE, and the more likely s/he would start a social venture. A concern about potential competition between businesses/charities and social enterprises arises as organizational ecologists argue that the increasing population density drives competition and thus leads to a decrease in the founding rate of new organizations (Lander & Heugens, 2017; Miller & Eden, 2006; Suddaby et al., 2017). However, we argue that conflict and competition are less likely to happen between SE and organizational forms of business and nonprofit. SE exists due to the market and government’s failure to meet social needs or solve social problems (Austin et al., 2006; Stephan et al., 2015). It identifies opportunities from societal problems that are not usually considered by commercial businesses, leaving space for SE development (Austin et al., 2006; Saebi et al., 2019). SE and nonprofit organizations are more likely to mutually be supportive of each other as both aim to address social problems and create social values (Battilana & Lee, 2014; Saebi et al., 2019). Therefore, we propose the following hypotheses that capture legitimacy spillovers concerning individuals’ engagement in SE:

Hypothesis 1a (H1a). New business density at the national level is positively associated with the likelihood of individuals engaging in SE.

Hypothesis 1b (H1b). Nonprofit organization density at the national level is positively associated with the likelihood of individuals engaging in SE.

3.2 Perceived SE density and SE engagement

As a new organizational form grows in size, it can also gain cognitive legitimacy through the intrapopulation process (Audia et al., 2006; Ruef, 2000; Xu et al., 2014). Here, the cognitive legitimacy of a given organizational form is based on the widespread knowledge about this form and therefore depends on the prevalence of the organizational form (Hannan & Freeman, 1984). When new organizational forms or practices become more prevalent within an area, there are more shared scripts and understandings about them; they further become more comprehensible and unquestioned to individuals (i.e., cognitive legitimacy). Therefore, the popularity of a new organizational practice increases its comprehensibility and is most commonly linked to its cognitive legitimacy (Bird & Wennberg, 2014; Husted et al., 2016; Schultz et al., 2014; Sine et al., 2005).

In line with the organizational ecology perspective, we expect that SEs’ density can also serve as a source of cognitive legitimacy for SE. After decades of development, SE has increasingly gained popularity in the world (Miller et al. 2012; Saebi et al., 2019; Short et al., 2009). As discussed above, a higher prevalence of an organizational form increases its comprehensibility (i.e., cognitive legitimacy) (Husted et al., 2016; Schultz et al., 2014). Accordingly, SE is more likely to be recognized when there is a higher density of SEs in their countries. Also, here, a concern may arise that population density may reflect competition, which further leads to the decrease or exit of new organizational forms (Lewis et al., 2021; Schultz et al., 2014; Xu et al., 2014). However, as organizational ecologists argued, density only causes fierce competition when an organizational form becomes fully taken-for-granted (Dobrev & Gotsopoulos, 2010; Lander & Heugens, 2017; Suddaby et al., 2017). Since SE is still a relatively recent phenomenon and becoming taken-for-granted is still the central concern of SE’s development (Bhatt et al., 2019; Ewald Kibler et al., 2018), SE’s density can better represent its cognitive legitimacy rather than competition. Thus, as the number of SEs grows, the density of SE contributes to its increased acceptance and taken-for-grantedness.

Institutionalists and organizational ecologists further find that the cognitive legitimacy of a new organizational practice, sourcing from its density, can enhance its adoption and implementation. As the density of a new organizational form increases, it becomes more widely accepted, and further, it has a much higher chance of being adopted by an individual. For instance, in the context of corporate social responsibility (CSR), Husted et al. (2016) show that the rising CSR density enhances its legitimacy, thus increasing a focal firm’s CSR engagement. We extend this logic to uncover the role of cognitive legitimacy at the micro level. We argue that the cognitive legitimacy of SE as perceived by an individual will enhance the likelihood of this individual engaging in SE. We focus on the perceived cognitive legitimacy of SE because the degree to which the legitimacy of a new organizational form determines an individual’s decision to engage in SE is dependent on individual perception (Miller et al., 2012; Nicholls, 2010; Scott & Lane, 2000). Bitektine and Haack (2015) also highlight the relevance of perceived cognitive legitimacy (via a mechanism of validity and collective judgment) in their multilevel model of legitimacy towards a certain observable action. Therefore, we propose that an individual’s perceptions of the social-enterprise form’s cognitive legitimacy can shape his/her choice to engage in SE. As we stated above, perceived cognitive legitimacy can be derived from residing in a region with a higher (perceived) density of social enterprise. We thus focus on individuals’ perceived SE density as an indicator of their perceived SE’s cognitive legitimacy and hypothesize:

Hypothesis 2 (H2). An individual’s perceived density of social enterprise is positively associated with the likelihood of individuals engaging in SE.

3.3 The interaction between perceived SE density and spillover effects

As aforementioned, individuals can increase their recognition and acceptance of SE through perceived SE density and legitimacy spillover effects, further leading to their entry into social entrepreneurial activity. However, these sources of cognitive legitimacy may not be strictly additive (Zimmerman & Zeitz, 2002). Institutional and organizational ecology theorists argue that individuals evaluate the legitimacy of a category differently as the category develops (Kuilman & Li, 2009; Taeuscher et al., 2021). The legitimacy spillover effect tends to play a more important role when the organizational form is relatively new. For instance, Zhao et al. (2018) find that an emerging category largely depends on conformity with other categories to obtain legitimacy during its early development. However, an organization’s legitimacy is characterized by a “range of acceptability” (Deephouse, 1999: 152). Once an organization reaches the audiences’ range of acceptability, it is perceived as legitimate and will gain only marginal benefits from additional legitimacy (Taeuscher et al., 2021). For example, Xu et al. (2014) show that an increasing density of new organizational forms can reduce the need for legitimacy transfer from other categories.

We attribute this change in the legitimacy evaluation of new organizational forms to different evaluative modes used by evaluators (Bitektine & Haack, 2015; Tost, 2011). Audiences evaluate the legitimacy of an organization using two different modes: either an evaluative or a passive mode (Bitektine & Haack, 2015; Jacqueminet & Durand, 2020; Tost, 2011). The evaluative mode is based on individual perception of the practice’s appropriateness, which requires more mental effort during the evaluation process. By contrast, in the passive mode, individuals either adopt collective beliefs and approval (i.e., validity) of the practice or quickly passively embrace practices that align with their expectations, which is effortless (Tost, 2011). The passive mode takes precedence unless there is a need or desire for the evaluative mode to intervene because individuals tend to conserve cognitive energy when evaluating (Tost, 2011). In addition, as a category has gained high legitimacy, evaluators face strong social pressure from its collective legitimacy, and those who make different evaluations are under pressure to suppress their deviant opinions (Centola et al., 2005; Zhu & Westphal, 2011). Thus, in this situation, individuals’ passive mode dominates the judgment process of an organization’s legitimacy, and the evaluative mode is suppressed (Bitektine & Haack, 2015). In contrast, when the population of a new category starts to grow, and there is a lack of perceived validity, the evaluative mode dominates the legitimacy judgment process (Tost, 2011).

As discussed above, the legitimacy spillover from both commercial business and nonprofit organizations to SE reflects an evaluative mode because it requires individuals to make efforts to understand the appropriateness of SE based on the two distinct categories it blends. By contrast, perceived SE density represents the perceived collective legitimacy of SE, thereby reflecting individuals’ evaluation in a passive mode (Bitektine & Haack, 2015). Therefore, in contexts with a perceived higher density of social ventures, individuals are more likely to accept SE passively to conserve energy (Lewis et al., 2021; Tost, 2011). As individuals become more exposed to SEs, the predictability and understandability of SEs increase, eliminating the need for individuals to interpret SE through other similar categories. That is, they rely less on the legitimacy spillover effects from business and nonprofit organizational forms. We, therefore, propose:

Hypothesis 3 (H3a). The positive legitimacy spillover effect of the business category on individual engagement in SE is weaker if the individual perceives a higher SE density.

Hypothesis 3 (H3b). The positive legitimacy spillover effect of the nonprofit category on individual engagement in SE is weaker if the individual perceives a higher SE density.

Figure 1 presents the conceptual framework combining Hypotheses 1–3 and showcases the multilevel nature of the mechanisms we propose for analyzing the cognitive legitimacy of SE.

Fig. 1
figure 1

Theoretical model

Full size image

4 Methods

4.1 Sample and data sources

To examine our model, we utilized a multilevel design incorporating individual-level (level 1) data nested within country-level data (level 2). To test our hypotheses, we combined multiple datasets. Firstly, our individual-level data were derived from the Adult Population Survey (APS), a sizeable population-representative survey conducted by the Global Entrepreneurship Monitor (GEM) in 2015 (Lepoutre et al., 2013; Stephan et al., 2015). The APS survey by GEM selects a random sample of at least 2000 individuals each year in every participating country, ensuring broad representation. These participants were interviewed either through phone or face-to-face interviews. For detailed information on data collection protocols, we refer to the GEM manual (Bosma et al., 2012).

Previous studies have recognized the GEM survey as a reliable data source, and numerous empirical cross-national entrepreneurship studies published in reputable academic journals have utilized GEM data for their analyses (Boudreaux et al., 2019; Kibler et al., 2018). In 2015, the survey collected responses from over 181,000 individuals across 60 countries. In addition, the 2015 GEM survey emphasized social entrepreneurship, incorporating specific questions that aimed to identify and capture individuals’ social entrepreneurial activity.

The data for our country-level variables were gathered from various sources, including the World Values Survey and the World Bank. To prevent any potential endogeneity issues between our predictors and the outcome, we lagged all country-level variables by at least one year. Once we merged the individual data with the aforementioned country-level data sources, our final sample encompassed at least 48,906 individuals from 22 countries, ensuring that all personal data and SE engagement information were accessible. The number of observations by country is presented in Table 1 in the Electronic Supplementary Material.

4.2 Variables and measures

4.2.1 Dependent variable at the individual level: SE

This study’s dependent variable focused on individuals’ engagement in social entrepreneurial activity. To identify social entrepreneurs, we employed a two-question criterion derived from previous studies (Saurav Pathak & Muralidharan, 2016, Pathak & Muralidharan, 2018; Sahasranamam & Nandakumar, 2020). Our measurement of SE engagement aligns with prior research on SE (Mair & Marti, 2006; Stephan et al., 2015). Participants were classified as social entrepreneurs if they responded affirmatively to both questions: “Are you, alone or with others, currently attempting to start or lead any activity with a social, environmental, or community objective?” and “In the past twelve months, have you taken any action to help initiate this activity, organization, or initiative?”. The dependent variable is treated as a binary variable, which is coded as 1 if individuals confirmed their current involvement in “trying to start,” “currently leading,” or both and had made efforts to commence the activity within the past twelve months. A code of 0 was assigned otherwise.

4.2.2 Independent variables at the country level

New business density

We measured the national-level new business activity using the total start-up activity in a country from the World Bank’s Entrepreneurship Survey. It is proxied by new firm density, which is the World Bank’s best-known indicator of total start-up activity. It represents the number of newly registered businesses (i.e., private, formal sector companies with limited liability) per 1000 working-age people (aged from 15 to 64), thus providing cross-national analysis data. Prior research has reported that this index has high validity and reliability (Carbonara et al., 2016; Herrera-Echeverri et al., 2014; Stenholm et al., 2013).

The density of nonprofit activity

According to prior studies, it is difficult to obtain homogeneous cross-national data on the nonprofit sector, such as the number of nonprofit organizations in a wide range of countries, donations, or employment within the third sector (Apinunmahakul & Devlin, 2008; Nissan et al., 2012; Saxton & Benson, 2005). Therefore, in this study, we measured the level of national nonprofit activity using a proxy variable—an average membership volume of nonprofit organizations of different types in a country. We obtained data from the World Values Survey Wave 6 (the 2010–2014 Wave). In this survey, individuals indicate whether they are active members of the following ten voluntary associations: (1) church or religious organization; (2) sport or recreational organization; (3) art, music, or educational organization; (4) labor union; (5) political party; (6) environmental organization; (7) professional association; (8) humanitarian or charitable organization; (9) consumer organization; and (10) self-help group, mutual aid group. The World Value Survey dataset further provides the country-level rate of all respondents that belong to each of the above associations. We computed the average rate across all ten associations. The score used in our analyses reflects the average percentage of individuals in each country’s sample actively involved in different nonprofit organizations.

4.2.3 Independent and moderating variable at the individual level: perception of SE density

As discussed in the hypotheses, individuals’ decision-making on engaging in SE relies on the legitimacy spillovers from the population-level business and nonprofit activity and individuals’ perception of SE legitimacy. To capture individuals’ perception of SE legitimacy, we used individual-level data relating to their perception of SE density in their country. We measure an individual’s perception of SE density as a dummy variable, which equals 1 if he/she responded in the affirmative that “in my country, you will often see businesses that primarily aim to solve social problems.” We obtained this data from the GEM APS database as well.

4.2.4 Control variables

We incorporated individual- and country-level control variables in line with prior multilevel empirical studies (Ewald Kibler et al., 2018; Stephan et al., 2015). At the individual level, we accounted for gender using a binary variable (female = 0, male = 1) (Estrin et al., 2013; Pathak & Muralidharan, 2018). Additionally, we included age as a categorical variable (Saurav Pathak & Muralidharan, 2023; Wei et al., 2023). Previous research has indicated a positive association between individuals’ education level and SE engagement (Estrin et al., 2013). Thus, we employed a seven-level education variable based on the GEM survey. Furthermore, we recognized the potential influence of work status and household income on SE activity (Saurav Pathak & Muralidharan, 2016; Sahasranamam & Nandakumar, 2020). We included work status as a binary variable, with a value of 1 indicating full-time or part-time employment. Similarly, household income was represented by a dummy variable, with a value of 1 signifying belonging to the middle- or higher-income group within their respective countries. Additionally, we accounted for individual socio-cognitive traits known to impact (social) entrepreneurial activity, such as perceived entrepreneurial self-efficacy, alertness to entrepreneurial opportunities, fear of failure, and individuals having a peer startup entrepreneur in the network (Boudreaux et al., 2019; Saurav Pathak & Muralidharan, 2016; Sahasranamam & Nandakumar, 2020). The specific questions used to measure these socio-cognitive traits are presented in Table 1.

Table 1 Variable descriptions
Full size table

At the country level, we included the logarithmic scale of GDP per capita for 2014 as a control variable to capture the potential country-level effects in our main regression model (Hoogendoorn, 2016). To conduct a series of robustness checks, we further considered three country-level variables (including GDP growth, unemployment, and postmaterialism) based on prior studies on SE (Hechavarría et al., 2023; Sahasranamam & Nandakumar, 2020). We obtained GDP per capita, GDP growth, and unemployment data from the World Bank. The indicator of postmaterialism is available from the World Value Survey, which was measured using a 4-item postmaterialism index (Inglehart, 1997). We used the data from the Wave 6 survey (period from 2010 to 2014). Postmaterialism reflects a country prioritizing prosocial attitudes, volunteering, and political activism (Stephan et al., 2015). We lagged for all country-level variables for at least one year. Table 1 shows all variables’ definitions and sources.

4.3 Method

We employed a multilevel modeling approach in our analysis due to the hierarchical nature of our dataset, which encompasses both national and individual levels. Individuals within the same country often share similar experiences, leading to comparable beliefs regarding SE. In multilevel modeling, observations within a nation are not considered independent of each other. In our study, it means that country-level factors influence the variation in assessing individuals’ willingness to engage in SE. Thus, for our main analyses, we utilized multilevel logit regression. This model offers several advantages compared to standard logit regression. Firstly, it helps overcome the ecological fallacy, which assumes that relationships observed at the aggregate level hold true at the individual level. We can avoid this fallacy by considering the multilevel structure (Jargowsky, 2005: 715). Secondly, the multilevel logit model accounts for the nonindependence of error terms, resulting in more reliable estimates compared to single-level regression (Estrin et al., 2013).

Consistent with the study conducted by Estrin et al. (2013), we assessed the statistical suitability of employing multilevel modeling in our analysis. Initially, we ran an empty multilevel regression model, which does not include any explanatory variables, to examine the significance of country effects (random intercepts). The findings revealed statistically significant country-level variances (p < 0.001) in explaining individuals’ engagement in social entrepreneurial activity. The residual interclass correlation (ICC1) indicated that 15.28% of the variation in SE engagement could be attributed to the country level, indicating a substantial proportion (Hox et al., 2017). To further evaluate the appropriateness of the multilevel modeling approach, we compared the ICC1 before and after incorporating macro-level control variables such as GDP per capita, business entry rate, and nonprofit activity. The results showed that the ICC1 decreased to 8.44% after accounting for these macro-level controls, suggesting that multilevel modeling remains suitable for our research purposes.

We utilized the variance inflation factor (VIF) method to assess the multicollinearity issue, as presented in Table A2 in the Electronic Supplementary Material. Our analysis indicated that multicollinearity was not a concern among our variables, as all VIF scores remained below the critical threshold of 5 (Hair, 2009).

5 Results

Within our study, Table 2 presents the descriptive statistics and correlation matrix for all variables utilized. We further examine our hypotheses by employing a series of logistic multilevel regression models, as our research aim was to explore whether population-level legitimacy can spill over to influence an individual’s SE decision-making, a binary variable following the Bernoulli distribution. Table 3 displays eight logit regression models used to test our hypotheses.

Table 2 Descriptive statistics and correlations
Full size table
Table 3 Multilevel logit regression results
Full size table

Model 1 incorporates all individual- and country-level control variables. Model 2 adds the entry rate of new firms, the first focal predictor, to test hypothesis 1a, and Model 3 adds nonprofit activity, our second focal predictor, to examine hypothesis 1b. To test hypothesis 2, Model 4 adds individuals’ perception of SE density as an independent variable. Model 5 includes all focal predictors. To test hypotheses 3a and 3b, Models 6–7 contain the interaction terms in a stepwise manner. We also incorporated both interaction terms in one model, as shown in Model 8. Finally, Models 9–16 provide a series of additional analyses to examine the robustness of our main results.

5.1 Main results

Before addressing the results linked to our hypotheses, we briefly discuss those control variables that predominantly confirm earlier findings. From Model 1, we observe that the probability of being engaged in SE differs among distinct age groups. Specifically, compared to the 18–24 age category, age this probability is lower for individuals aged 25–34 (β = −0.197, p < 0.01) and 35–44 (β = −0.185, p < 0.05), while it is larger for those in the 55–64 age range (β = 0.175, p < 0.05). Model 1 also finds that people are more likely to pursue SE within groups characterized by higher levels of educational attainment. Individuals who are either fully or part-time employed are also more inclined to be involved in SE (β = 0.499, p < 0.01). Household income also shows a positive relationship with involvement in SE (β = 0.16, p < 0.01). Consistent with prior findings of e.g. Boudreaux et al. (2019) and Sahasranamam and Nandakumar (2020), Model 1 indicates a strong association between individual socio-cognitive entrepreneurial traits and self-reported SE activity. Specifically, we find individuals’ having a peer startup entrepreneur in their network (β = 0.68, p < 0.01), entrepreneurial self-efficacy (β = 0.543, p < 0.01), and alertness to entrepreneurial opportunity (β = 0.349, p < 0.01) to be positively associated with the probability to be engaged in SE, while fear of failure presents a negative relation (β = −0.162, p < 0.01).

Turning to our hypotheses, we proposed a positive relationship between the entry rate of new businesses at the national level and individuals’ SE entry in Hypothesis 1a. In Model 2, we find the entry rate of new businesses was positively associated with an individual’s SE entry, and the relationship was statistically significant at the 5 percent level. The coefficient of New business density is 0.085, indicating that if the number of new businesses per 1000 residents increases by 1, the log odds ratio of SE entry increases by 0.085. From Model 3, we also support Hypothesis 1b, as a country’s nonprofit activity significantly positively affects an individual’s SE entry (β = 0.074, p < 0.05). Model 4 supports Hypothesis 2 that an individual’s perceived density of social enterprise is significantly and positively associated with the likelihood of individuals engaging in SE (β = 0.214, p < 0.01). We still found positive and robust relationships when all focal predictors were included in the same model (Model 5). The average marginal effects of new business density and nonprofit activity are 0.004 and 0.003, suggesting that the average probability of an individual engaging in SE increases by 0.004 and 0.003, respectively, if new business density and nonprofit activity by one unit.

The interaction models further indicate that an individual’s perception of SE density negatively moderates the effects of legitimacy spillovers on an individual’s SE engagement. Specifically, from Model 6, we find a significant negative interaction between the entry rate of new businesses at the national level and individuals’ perception of SE density (β = −0.028, p < 0.05), which supports our Hypothesis 3a. Finally, in Model 7, we note that an individual’s perception of SE density also has a significant negative moderating effect on the influence of a nation’s nonprofit activity on individuals’ SE engagement (β = −0.045, p < 0.01), providing support for our Hypothesis 3b as well. When we included both interaction terms into one model (Model 8), the results still supported Hypothesis 3a and 3b.

We plot the notable moderating effects that depict the average predicted probability of individuals’ involvement in SE across different levels of perceived density of SEs while considering varying densities of both new business and nonprofit organizations. In Fig. 2, the vertical axis represents the probability of engaging in social entrepreneurship, while the horizontal axis represents the density of newly established businesses within a country. The results in Fig. 2 indicate that as the density of new businesses increases, both people who perceived SEs’ density and those who did not are more likely to engage in social entrepreneurship than their counterparts in societies with a lower density of new business. In addition, the legitimacy spillover effect from the new business on SEs becomes smaller when individuals perceive a density of SEs in their countries. We plot the graphs for the significant interaction effect between nonprofit organizations and perceived SE density in social entrepreneurship (Fig. 3). It shows that people with a perception of SE density are less likely to depend on the legitimacy spillover effect from the nonprofit organizations for deciding to start a social venture. Overall, all figures suggest that the legitimacy spillover effects become smaller in predicting individuals’ SE entry when they perceive a density of SEs in their countries.

Fig. 2
figure 2

Interaction between country-level new business density and individual-level perceived SEs density in social entrepreneurship

Full size image
Fig. 3
figure 3

Interaction between country-level nonprofit organizations and individual-level perceived SEs density in social entrepreneurship

Full size image

5.2 Robustness tests

We then conducted a series of robustness checks, as shown in Table 4. First, all the positive effects of our focal predictors and the interaction effects were replicated by removing alertness to entrepreneurial opportunity as an individual-level control variable. The results support the robustness of our findings. Then, we added GDP growth and unemployment as country-level control variables separately. The results provide evidence of the robustness of our findings. Finally, we used postmaterialism from the World Value Survey as an alternative measure for one of our predictors, the national-level nonprofit activity. We repeated all the analyses. The results are consistent with our original findings, as presented before, except for a less robust moderating effect of perceived SE density on the relationship between postmaterialism and individuals’ SE entry (Model 16). Apart from the above robustness tests, we further run our analysis by adopting an alternative measurement for our dependent variable (i.e., individuals’ engagement in SE) and using the lagged density of business by two years, respectively. The results still support our hypotheses (see Table A3 and Table A4 in the Electronic Supplementary Material).

Table 4 Robustness check
Full size table

6 Discussion and conclusions

This study explores the pathways of how SE as an organizational form may obtain cognitive legitimacy. Drawing on legitimacy theory and organizational ecology perspectives, we find that SE can gain cognitive legitimacy by means of legitimacy spillover effects from both commercial business and nonprofit categories (Kostova & Zaheer, 1999; Kuilman & Li, 2009) and by the perceived prevalence of social enterprises (Hannan & Freeman, 1984). We also argue that these two sources of cognitive legitimacy are not strictly addictive. We find legitimacy spillover plays a less vital role when individuals perceive a higher prevalence of SE activity. The empirical results suggest both the density of business and nonprofit activity in a country are positively associated with individuals’ social entrepreneurial entry; moreover, this positive association is diminished when individuals perceive SEs to be prevalent in their country.

6.1 Theoretical implications

This paper makes contributions to research on SE, organizational ecology, and hybrid organizations. First, we contribute to SE literature by identifying its sources of cognitive legitimacy. Prior SE research initially considered studying SE’s cognitive legitimacy to be theoretically unwarranted (Dart, 2004) and has, accordingly, paid limited attention to SE’s cognitive legitimacy (Kibler et al., 2018; Miller et al., 2012). This paper adds to the understanding of SE’s cognitive legitimacy by identifying two sources at different levels of analysis: legitimacy spillovers from business and nonprofit organizations (macro level) and SE prevalence perceived by individuals (micro level). We propose that the macro-level legitimacy spillover effects of both business and nonprofit categories should be considered in conjunction with the perceived presence of SE at the micro level in order to assess SE’s cognitive legitimacy, but in different ways: the perceived density of SE enables individuals to passively consider SE as comprehensible, while the macro-level legitimacy (i.e., density) of business and nonprofit organizations assist individuals to actively evaluate the comprehensibility of SE (i.e., legitimacy spillover). Our results further suggest that the legitimacy spillover effects become weaker if individuals perceive a higher SE density, thus providing empirical support to previous theories about how individuals evaluate legitimacy in either evaluative or passive modes and how the passive mode dominates the evaluation process (Bitektine & Haack, 2015; Tost, 2011).

Second, we contribute to the organizational ecology literature by leveraging a multilevel theory of legitimacy proposed by Bitektine and Haack (2015). Organizational ecologists have long discussed the prevalence of an organizational form by investigating interpopulation (i.e., legitimacy spillover) and intrapopulation processes (i.e., the accumulated legitimacy through their own density) (Kuilman & Li, 2009; Lewis et al., 2021; Xu et al., 2014; Zhao et al., 2018). They mainly focus on understanding both processes at the macro level (Dobrev et al., 2006; Hannan & Carroll, 1992; Kuilman & Li, 2009; Li et al., 2007; Zavyalova et al., 2012). Our multilevel model explains how interpopulation and intrapopulation processes occur through a cross-level mechanism: the macro-level legitimacy of business and nonprofit organizations, as well as the perceived SE density, affect individuals’ comprehension of SE, thereby influencing their engagement in SE. In addition, earlier organizational ecologists mainly focus on the legitimacy spillovers within categories (e.g., from a category to its subcategory) (Dobrev et al., 2006; Kuilman & Li, 2009; Li et al., 2007; Zavyalova et al., 2012). We extend previous research by focusing on how legitimacy spillovers occur across categories. That is, we find that social enterprises, which straddle contradicting categories of business and nonprofits, can receive legitimacy spillovers from both categories.

Finally, this paper contributes to the current body of literature on hybrid organizations. It responds to recent calls for more symmetrical theorization of hybridity to better understand its positive and negative effects (Wry et al., 2014). Scholars usually find that mixed identities, organization forms, and institutional logics in hybrid organizations could confuse what ‘type’ they belong to, which leads to tensions and may inhibit them in obtaining and maintaining legitimacy (Greenwood et al., 2011; Kostova & Zaheer, 1999). Instead, our results suggest that SEs may also benefit from their hybrid nature, acquiring legitimacy over time through the spillover effects from both the established business and nonprofit categories that it straddles. It also holds the potential to contribute to a better understanding of the emergence and legitimacy of other types of hybrid ventures, such as sustainable and environmental entrepreneurship.

6.2 Implications for practice

The findings of this study may hold significant implications for policymakers seeking to promote social entrepreneurial activity. Our results indicate that the presence of new businesses and nonprofit initiatives within a country can positively influence individuals’ engagement in SE. Therefore, facilitating the establishment of new startups and supporting nonprofit activities in a country or region may also be conducive to the growth of social entrepreneurial activity. Policymakers are encouraged to enhance government support and foster a conducive business ecosystem for startups and nonprofit organizations while developing policies to foster SE. This approach not only benefits the expansion of startups and nonprofits but also serves as a catalyst for attracting more individuals to participate in SE. Encouragement can be manifested through various means, such as providing incentives to signal benefits (e.g., financial support or direct subsidies) and establishing supportive regulative institutional frameworks (Stenholm et al., 2013).

Policymakers should be aware that the positive legitimacy spillover effects on individuals’ SE engagement tend to diminish when individuals perceive a greater density of SE in their countries. In other words, the initial legitimation for SE activity comes from its mixed business and nonprofit categories with well-established legitimacy when the number of social enterprises is small. However, as the SE population grows and accumulates constitutive legitimacy that is most commonly related to the proliferation of an organizational population, individuals who perceive a higher social recognition for SE in their country are less likely to rely on legitimacy spillover effects in their decision to engage in SE. Therefore, policymakers are advised to formulate a plan for SE’s long-term development, taking into account legitimation over time due to legitimacy spillovers. For example, depending on the existing legal structures that may be enabling or hindering SE, granting sociopolitical legitimacy to SE through the enactment of legislation and laws can be an effective approach, as previous research has highlighted the importance of government endorsements as a source of legitimacy for new organizations (Tost, 2011).

6.3 Limitations and future research

While our study contributes to understanding legitimacy spillovers for social entrepreneurial activity, we acknowledge several limitations. This study employed a population-representative sample from 22 diverse countries to test our multilevel hypotheses. It is important to note that the cross-sectional nature of our dataset limits our ability to establish strong causal relationships based on our analyses. Future research could employ alternative methodologies, such as experimental or longitudinal research designs, to provide further insights into the causal relationships under investigation.

Another limitation of our study is that our measure of SE relies on a single indicator that reflects the overall SE activity. However, Zahra et al. (2009) categorized social entrepreneurs into three distinct types: social bricoleur, social constructionist, and social engineer. These different types of social entrepreneurs exhibit variations in their ability to identify opportunities and assemble the necessary resources to pursue those opportunities. Hence, it becomes crucial to investigate how legitimacy spillovers may influence each type of social entrepreneur differently. Unfortunately, the available GEM dataset does not provide the means to differentiate between these types of social entrepreneurs. Future research should address this limitation by considering the diverse categories of social entrepreneurs to explore the potential varying effects of legitimacy spillovers across these different types.

Third, organizational scholars have recognized multiple legitimacy types (i.e., cognitive, normative, and regulatory legitimacy) that are important for organizations (Bitektine, 2011; Scott, 1995). However, in this study, we specifically explored the spillover effects of constitutive legitimacy on SE activity. Future research may investigate how other types of legitimacy spill over to SE activity. For instance, researchers could examine whether granting sociopolitical legitimacy to business or nonprofit sectors leads to higher social recognition for SE activity in the future.

Finally, while this paper has revealed the positive legitimacy transfers across categories (from business and nonprofit sectors to SE), the legitimacy of SE could also be negatively affected by its cognitively related categories. Prior studies have suggested that firms can suffer from negative spillovers when members of their industries engage in wrongdoing (Barnett & King, 2008; Zavyalova et al., 2012), and negative spillovers may have a stronger influence on legitimacy than positive spillovers (Haack et al., 2014; Kostova & Zaheer, 1999). Furthermore, this paper focuses on a so-called ‘top-down’ vertical spillover across categories. It is also possible to explore how the overall legitimacy of the “business” and “nonprofit” categories could be affected by SE activity, leading to ‘bottom-up’ legitimacy spillovers. Therefore, it may be fruitful to explore such bottom-up legitimacy spillovers, which can take place simultaneously as “top-down spillovers” in future studies.