Chance to Shine or Heading for Disaster? – Exploring German Social Enterprises’ Stakeholder Communication Facing COVID-19-Induced Lockdowns

Abstract

During COVID-19 pandemic, lockdowns were imposed to contain infections. For social enterprises (SEs), that apply entrepreneurial means to fulfil their social missions, research suggests lockdown-related chances and risks. However, no insights on SE stakeholder communication facing COVID-19-induced lockdowns exist. Our work investigates SEs’ communicational framings for two successive lockdowns based on 121 German SEs. Analysing Facebook data with a mixed methods approach, we show that, after both lockdowns, chance-oriented framings dominated. However, they decreased after the second lockdown. Despite a limited generalisability and a potential positivity bias in our data, we offer notable insights regarding SE stakeholder crisis communication.

Keywords:

• Social entrepreneurship
• covid-19
• lockdown
• stakeholder communication
• crisis
• mixed methods

Introduction

Entrepreneurship is a driver of innovativeness, economic competitiveness, and generates financial wealth for the successful entrepreneur and his/her stakeholders (van Praag and Versloot 2007). While, throughout the history of entrepreneurship research, commercial perspectives have been dominant (Murphy, Liao, and Welsch 2006), in the last four decades, entrepreneurial behaviour was increasingly realised as a means to address pressing social problems like poverty, disparity or social marginalisation (Austin, Stevenson, and Wei-Skillern 2006; Saebi, Foss, and Linder 2019;Young 1983;). This new form of entrepreneurship is coined social entrepreneurship and driven by a social mission. However, in contrast to traditional non-profit organisations, social enterprises (SEs) build on an elaborated business plan and generate their own income (Kruse and Rosing 2023; Kruse, Wach, and Wegge 2021; Stephan et al. 2016). Therefore, SEs are considered ‘hybrid’, as they create social and financial value. This yields several chances like independence from donors’ good will or political influence (Dupuy, Ron, and Prakash 2016) and numerous SE success stories are reported in literature (cf. Perrini, Vurro, and Costanzo (2010); Thompson and Doherty (2006)). However, running an SE also involves notable challenges like a higher risk of failure compared to commercial enterprises (McCaffrey 2018) or mission drifts, i.e. the prioritisation of commercial over social goals over time which could be caused by crises threatening enterprise survival (Grimes, Williams, and Zhao 2019).

Probably the most massive and large-scale crisis economies and societies worldwide had been confronted with since World War II was the outbreak of the previously unknown severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and the subsequent COVID-19 pandemic (Le and Nguyen 2021; Singh et al. 2021;Winston 2020). In the face of a steeply rising infection curve and death toll and as there was neither a working medication nor vaccination available, several governments took strict measures to prevent further infections. Acting on the maxim to reduce social contacts to a minimum, typical actions encompassed social distancing and the closure of schools, restaurants, and almost all shops. These measures were coined ‘lockdowns’ (Koh 2020). COVID-19-induced lockdowns are overwhelmingly judged as a serious threat for enterprise survival due to their detrimental effects on sales and financial liquidity (see e.g. Kuckertz et al. (2020) for an overview). However, some scholars also see chances for enterprises, e.g. the possibility to address lockdown-related social problems like marginalisation, loneliness, and rising inequality (Schippers (2020) for a summary) through innovative entrepreneurial means (Bacq et al. 2020; Bacq and Lumpkin 2021; Scheidgen et al. 2021; Weaver 2020). Due to their social missions, SEs seem particularly suitable to deliver such solutions, while, from a financial perspective, SEs are hit even harder by COVID-19-induced lockdowns than commercial enterprises. As social entrepreneurs deliberately accept lower revenues to fulfil their social mission, they are more susceptible to financial problems. Consequently, there is a notably higher risk to run out of liquidity and eventually fail (Belitski et al. 2022; Weaver and Blakey 2022).

Despite notable insights on SE reactions to the COVID-19 pandemic (see e.g. Kročil, Müller, and Kubátová (2023); Loukopoulos and Papadimitriou (2022); Ruiz-Rosa, Gutiérrez-Taño, and García-Rodríguez (2020); Weaver (2020)), reviewing pertinent literature yields two major shortcomings. First, there is an almost exclusive focus on the social entrepreneur and his/her perceptions and reactions. While this mirrors a general tendency in the field (Brattström and Wennberg 2022; Maclean, Harvey, and Gordon 2013), it neglects the innate complexity of the SE phenomenon (Saebi, Foss, and Linder 2019), e.g. the crucial role of SE stakeholders. Stakeholders, defined as constituencies or individuals affecting SE activities and value generation (Freeman 2004), are key for enterprise viability (Coombs and Holladay 2015; Post, Preston, and Sachs 2002). Considering COVID-19-induced chance-risk-ambiguity and uncertainty, SE-stakeholder communication can be a valuable tool to sustain stakeholder trust and commitment (Coombs 2007). Second, previous studies largely neglect that, in most countries, successive lockdowns occurred due to the wave-like nature of the pandemic (Dutta 2022). While government measures remained relatively consistent, one can presume that, following the experiences made during the first lockdown, there was a learning curve resulting in a change of SE reactions, for example regarding their stakeholder communication.

Addressing these two apparent gaps, we explore (i) the communicational framings SEs employed in their stakeholder communication related to COVID-19-induced lockdowns and (ii) potential changes comparing two successive lockdowns. We do so by combining qualitative (content analysis) and quantitative (Kruskal Wallis tests and linear mixed models) methodologies in a mixed methods approach analysing Facebook data of 121 German SEs. Thereby, our study makes original contributions by (i) taking a largely neglected stakeholder communication perspective, (ii) examining the application of cognitive framings in real social media communication, and (iii) deriving notable implications for future research and practitioners.

After conceptualising social entrepreneurship as a new form of entrepreneurship, we reflect on COVID-19-related SE-chances and risks, highlight the importance of stakeholder communication in times of crisis, and derive our research questions. Afterwards, we summarise governmental reactions to the pandemic in Germany, describe the choice and acquisition of our sample, outline our data sources and methods, and present the results. Finally, the results are discussed and implications and limitations reported.

Theoretical background

Social entrepreneurship as a new form of entrepreneurship

Throughout the history of entrepreneurship research, that dates back to the mid-18^th century and first essays by the Irish economist Cantillon (1756), entrepreneurial activity was primarily seen as a means to generate financial wealth for the entrepreneur and his/her stakeholders (Murphy, Liao, and Welsch 2006). In the 1980s, however, Young (1983) pioneered the idea of employing entrepreneurial behaviour for the alleviation of social problems like poverty or social marginalisation. Thereby he introduced the concept of social entrepreneurship. Social entrepreneurs combine a social mission as central driver of their actions with entrepreneurial behaviour generating their own income (Austin, Stevenson, and Wei-Skillern 2006). Thus, SEs are referred to as ‘hybrid enterprises’ that range between socially-minded non-profit organisations and commercially-minded enterprises (Kruse and Rosing 2023; Tracey and Phillips 2007). Whereas also commercial enterprises increasingly engage in social value creation, for example through corporate social responsibility (CSR; see Aguinis and Glavas (2012) for an overview), the core difference lies in its purpose. For SEs, social value creation is the fundamental driver of entrepreneurial activity while, for commercial enterprises, CSR fulfils rather instrumental purposes like attracting new, socially-minded customers (Justo et al. 2010; Stephan et al. 2016). In our research, we follow Kruse (2020a) and define an SE as ‘an enterprise whose business model is to address unmet socio-economic needs in communities in an innovative and financially sustainable way by creating social value and generating revenue for the enterprise and its stakeholders’ (p. 644).

Chances and risks for social enterprises facing COVID-19

Reviewing scholarly literature, several SE success stories can be found (Thompson and Doherty (2006) for an exemplary overview). To illustrate, the Italian SE San Patrignano is dedicated to the rehabilitation of former prisoners and drug addicts by offering work as bakers or dog trainers. The goods and services offered by San Patrignano are sold, generate income, and thereby secure enterprise survival (Perrini, Vurro, and Costanzo 2010). However, notwithstanding the clear chances of SE-activities to create financially independent and sustainable social value for its beneficiaries (Kruse, Chipeta, and Venter 2023), social entrepreneurial activity bears enormous challenges. Generally, social entrepreneurs accept lower revenues and income for the sake of their social missions. This makes them particularly vulnerable to crises and external shocks like COVID-19-induced lockdowns. As their financial buffer is, on average, lower compared to commercial enterprises, they have a higher risk of suffering from illiquidity and, ultimately, fail (Belitski et al. 2022; Weaver 2020; Weaver and Blakey 2022). Exploring Czech social entrepreneurs’ reactions to COVID-19-induced lockdowns based on in-depth interviews, Kročil, Müller, and Kubátová (2023) found that some social entrepreneurs felt ‘paralyzed and unable to take the necessary actions to overcome the crisis’ (p. 159).

While, the social mission can be the financial Achilles heel of SEs during crises, some scholars argue that it can be a genuine chance during the pandemic. As lockdown measures exacerbated social problems like marginalisation and inequality (cf. Schippers (2020)), SEs could have the chance to extend their operational range and, thereby, increase their social value creation activities (Bacq and Lumpkin 2021). Bacq et al. (2020) showed that, in the face of COVID-19 threats, creativity in the SE-context can grow. Examining ten Greek SEs, Loukopoulos and Papadimitriou (2022) found that some SEs saw the pandemic as a chance to benefit from ‘through expanding services, building new interorganizational collaborations and serving additional beneficiaries’ (p.541).

To conclude, current scholarly insights on SEs’ reactions to the pandemic and COVID-19-induced lockdowns yield a high ambiguity. On the one hand, lockdowns pose an enormous financial risk and threaten SEs’ financial viability. On the other hand, lockdown-related social problems are a chance to serve more beneficiaries and increase social value creation activities. As a result, one can presume that, also in SE-stakeholder communication, lockdown-related chances and risks play an important role.

(Social) enterprise stakeholder communication in times of crisis

Similar to commercial entrepreneurship, also in social entrepreneurship one can remark a ‘heroic bias’, i.e. an almost exclusive focus on the individual social entrepreneur (Brattström and Wennberg 2022). However, this overestimates the entrepreneur’s role and underestimates the importance of stakeholders (Dorado et al. 2022; Saebi, Foss, and Linder 2019). According to Post, Preston, and Sachs (2002), stakeholders (broadly defined) comprise ‘individuals and constituencies that contribute, either voluntarily or involuntarily, to its wealth-creating capacity and activities, and who are therefore its potential beneficiaries and/or risk bearers’ (p.7). This definition highlights the importance of stakeholders for (social) enterprises, as they have the potential to sustainably affect (social) enterprise performance, for example by supplying critical resources (Freeman 2004). Thus, as summarised by Stakeholder-Agency Theory, the interaction of (social) entrepreneurs and stakeholders through means of communication and action are key to sustained enterprise success (Davis et al. 2021; Hill and Jones 1992).

Particularly, in crises, defined by Coombs (2015) as ‘the perception of an unpredictable event that threatens important expectancies of stakeholders […] and can seriously impact an organization’s performance and generate negative outcomes’ (p. 3), stakeholder communication is fundamental to influence and shape stakeholder perception of the crisis and, thereby, prevent (reputational) damage (Coombs and Holladay 2015). Drawing from Situational Crisis Communication Theory (SCCT) as one of the most comprehensive and influential theories in the field of corporate post-crises communication (Coombs 2007), framing is probably the most powerful tool to affect stakeholders’ crisis perception.

Framing refers to the selective and goal-directed application of communicational cues and contents to guide stakeholders’ attention (Druckman 2001). By setting a communicational frame, e.g. when stressing the innate chances in a crisis, communicators try to affect the formation of recipients’ opinions and, ultimately, increase the probability that judgements are made in favour of the communicator (Coombs 2007). While framing is a universally applicable strategy to influence stakeholder perceptions of crises, different framing strategies exist. These strategies are more or less suitable depending on the form of a crisis (Coombs and Holladay (2015) for an overview). According to SCCT, COVID-19-induced lockdowns are a victim crisis, i.e. an externally-triggered crisis that could neither be foreseen nor controlled by the enterprise. Consequently, the enterprise itself is a victim. As a response to victim crises, two framing strategies seem feasible. First, enterprises could emphasise their victimhood and stress potential risks and problems currently encountered. By doing so, enterprises try to trigger compassion amongst stakeholders and, thereby, secure stakeholder support. This framing strategy is referred to as compassion framing (Coombs 1999). Second, enterprises could frame the crisis as an opportunity and chance to benefit from. This way, enterprises emphasise their willingness to remain in the driving seat, (pro-)actively navigate their enterprise through the crisis, and become stronger in the future. In other words, enterprises frame their reaction as ‘making a virtue out of necessity’ (cf. Forde et al. (2021); Semonella et al. (2022)), henceforth, referred to as MAVOON framing.

Through an SE lens, both framing strategies can be considered feasible reactions to lockdowns. On the one hand, compassion can be triggered by focusing on the devastating financial consequences of shop-closures and other lockdown restrictions. On the other hand, emphasising social problems resulting from lockdowns and ones own capacity to help can frame lockdowns as a chance to extend operational fields in the future. In fact, previous research indicates that social entrepreneurs perceive chances (Bacq et al. 2020; Loukopoulos and Papadimitriou 2022) and risks (Kročil, Müller, and Kubátová 2023; Weaver and Blakey 2022) facing COVID-19-induced lockdowns. However, it remains unclear how they frame these perceptions in their real stakeholder communication. Our current knowledge base consists of conceptual work or interviews of social entrepreneurs with scientists who are no distinct SE-stakeholders. Thus, we explore the application of compassion and MAVOON framing in real SE stakeholder communication and examine the following research question (RQ):

RQ₁: To which extent do social enterprises apply compassion and making-a-virtue-out-of-necessity framing in their stakeholder communication after COVID-19-induced lockdowns?

Another open question is, whether, in SE-stakeholder communication, framing changed during successive lockdowns. Due to the wave-like nature of the pandemic, several governments imposed more than one lockdown (Dutta 2022). In contrast to the first lockdown, a reduction of uncertainty on both sides (SEs and stakeholders) occurred over time and it seems plausible that this is associated with their (communicative) behaviours. On the person level, studies state a behavioural adaptation of individuals facing successive lockdowns (see e.g. Sullivan et al. (2021)). However, so far, similar insights for SEs and their stakeholder communication do not exist. Thus, we derive the second research question as follows:

RQ₂: Are there changes in the application of compassion and making-a-virtue-out-of-necessity framing in social enterprise-stakeholder communication comparing successive COVID-19-induced lockdowns?

While lockdowns were imposed in several countries worldwide, notable differences, for example, regarding duration and strictness existed (cf. Koh (2020)). For the sake of clarity, the current study focuses on one country only. We chose Germany due to three reasons. First, Germany imposed two successive nationwide lockdowns with very similar measures. This corresponds to the scope of our research. Second, due to the existence of a German social entrepreneurship network (‘Social Entrepreneurship Netzwerk Deutschland’), SE-activity is recoded and a comprehensive database of SEs exists (Kruse 2022). Third, as the identification and correct coding of framings requires sophisticated language skills, we, as German native speakers, limited our scope to German posts for the sake of coding reliability.

Materials and methods

COVID-19-induced lockdowns in Germany – a brief overview

After registering the first officially confirmed COVID-19 case in Germany in late January 2020, infection numbers quickly saw an exponential rise and the first death following a COVID-19 infection was registered in early March. On 22nd March, the German government imposed the first lockdown to prevent infections from spreading further (Grote et al. 2021, Niedenhoff and Orth 2021). The measures taken included unprecedented restrictions regarding individual mobility and social life, as social contacts should be reduced to a minimum. Following this maxim, schools, most offices, restaurants, and almost all shops had to close. As it was unclear how fast the lockdown would lead to a falling infection curve, no definite end-date was set. Following a decrease in infections, the first lockdown was eventually ended on 4th May. While, during summer and early fall, infections remained steadily low, a steep rise in course of the second COVID-19 wave occurred in late 2020 resulting in the second lockdown from 16th December to early March 2021. Subsequently, the German government tried to avoid nationwide lockdowns and focused more on the containment of COVID-19 infections in hotspots (e.g. Federal States or communities) with particularly high infection rates. This led to a more scattered landscape of different regulations throughout the country. Generally, and in the light of more and more people recovering from COVID-19 and/or getting vaccinated, measures were eased for the majority of Germans and the last COVID-19-related restrictions expired in early April 2023.

For our research, the first two lockdowns imposed on 22nd March and 16th December 2020 are particularly relevant, as they (i) were the only nationwide lockdowns in Germany and (ii) featured very similar measures. While we acknowledge that, despite uniform lockdown measures, regional differences in infection rates existed, our focus lies on the communicational reactions to lockdowns and not COVID-19 infection rates. Thus, we consider these two lockdowns suitable for comparative analyses (cf. RQ₂).

Data source

A wide and growing variety of social media platforms exists and an increasing number of scholars sees potential in using them as a data source (Ham et al. 2018). One central advantage of social media data is that, compared to ‘traditional’ survey or interview methodologies, ‘natural’ communication can be assessed (Kaiser and Kuckertz 2023). Other notable assets of social media data encompass an easy access, public availability, and automatic time stamps making it easy to link posts with specific events like the imposition of a lockdown (Vitak 2017). Furthermore, in the business context, social media gained traction as a tool to communicate with stakeholders and attract customers (Nikbin et al. 2022; Sivertzen, Nilsen, and Olafsen 2013), especially for small and medium-sized enterprises (Ali, Balta, and Papadopoulos 2014; Brooks, Heffner, and Henderson 2014). As our study goal was to examine real SE-stakeholder communication reacting to COVID-19-induced lockdowns, we decided on acquiring data from a social media platform.

Comparing Instagram, X (formerly Twitter), and Facebook as three of the most commonly used platforms, we decided on Facebook as our data source due to two main reasons. First, on average, Facebook posts contain more text than image-based Instagram posts. In contrast to texts, images can be ambiguous and hard to code. Correspondingly, Instagram studies usually employ a survey methodology assessing user attitudes or behaviour as a reaction to Instagram posts (e.g. Asdecker (2022)). However, this this does not correspond to our research questions. Second, while Facebook and X are both text-based, Facebook was the dominating platform in our sample. Only four SEs had to be excluded, as they did not possess a Facebook account (cf. Figure 1) while 18 SEs did not have an X account. This corresponds to the finding by Brailovskaia, Schillack, and Margraf (2018) who could show that Facebook is the most frequently used social media platform in Germany.

Figure 1. Flow chart depicting dropouts on the way to the final study sample. Note. SEND: Social Enterprise Network Germany.

Sampling procedure and sample description

For the acquisition of social enterprises, we contacted Germany′s largest social entrepreneurship network SEND (‘Social Entrepreneurship Netzwerk Deutschland’). In total, the network has more than 800 members, all small and medium-sized enterprises, and represents SE-related interests in business and politics (Kruse 2022). To determine an adequate sample size for our coding, we applied a three-step procedure. First, we conducted an a-priori power-analysis using GPower (Faul et al. 2007). As, to the best of our knowledge, this is the first study evaluating communicational framings in the SE-context, we could not base our effect size considerations on similar studies. Thus, we reviewed psychological literature more broadly. This yielded an average effect size of r = 0.21 (Fraley and Marks 2007; Funder and Ozer 2020; Richard, Bond, and Stokes-Zoota 2003). Due to the severity of COVID-19-induced lockdowns, we expected above-average effect sizes and calculated two separate power analyses for effect sizes of r = 0.25 and r = 0.30. As Figure 2 shows, this yielded a sample size range of 120 ≤ N ≤ 175. Second, expecting dropout rates up to 30%, we randomly drew 180 SEs from the entire database. We did so by numbering all 800 SEs included in the initial database. We then randomly generated 180 numbers (range: 1 to 800) and included the enterprises matching the numbers. Third, reviewing the enterprises drawn in more detail yielded some exclusions of enterprises not suitable for our study (cf. Figure 1). Three SEs appeared not to be operating anymore. 35 SEs were not affected by lockdown measures (e.g. SEs selling food could remain open) or had no Facebook profile. Four SEs had inaccessible or private Facebook profiles and 17 SEs did not make posts in our period of interest after at least one of the two lockdowns. This resulted in a final sample of 121 SEs.

Figure 2. Power analysis results for effect sizes r = 0.25 and r = 0.30 calculated with GPower (Faul et al. 2007).

SEs in our sample operated in seven different sectors (Table 1) and were situated in 13 of 16 German Federal States (Table 2). While this yields an uneven geographical distribution, our sample mirrors the German SE-landscape in which the majority of SEs can be found in urban regions (e.g. Berlin, Hamburg) and the economically strongest and most populous Federal States of Bavaria and North Rhine-Westphalia (Hoffmann et al. 2021). Regarding enterprise age, 17.35% were younger than 2 years, 25.62% were between 2 and 5 years old, and 57.02% older than 5 years. The mean number of Facebook subscribers per SE was 15,891.

Table 1. Sector distribution of social enterprises entering our analyses (N = 121).

Sector	n	%
Business services	18	14.9
Community- and social services	29	24.0
Education	24	19.8
Financial services	12	9.9
Health services	12	9.9
Agriculture, environment, and energy	17	14.0
Others	9	7.4
Total (N)	121	100^a

Note. ^aDue to rounding, percentages do not add up to exactly 100%.

Table 2. Federal state distribution of social enterprises entering our analyses (N = 121).

Federal state	n	%	Federal state	n	%
Baden-Wuerttemberg	10	8.3	Lower Saxony	2	1.7
Bavaria	20	16.5	North Rhine-Westphalia	16	13.2
Berlin	37	30.6	Rhineland-Palatinate	1	0.8
Brandenburg	4	3.3	Saarland	2	1.7
Bremen	3	2.5	Saxony	2	1.7
Hamburg	14	11.6	Schleswig-Holstein	2	1.7
Hesse	8	6.6
			Total (N)	121	100^a

Note. ^aDue to rounding, percentages do not add up to exactly 100%.; no social enterprises in Saxony-Anhalt, Thuringia, and Mecklenburg-Western Pomerania featured in our sample.

Qualitative analyses

In the first step, all Facebook posts published after the government announcement of the first and second lockdown were collected in an MS Excel sheet. Informed by previous event studies in the business context examining reactions to external shocks in general (Elad 2017) and COVID-19 in particular (Gu, Zhang, and Cheng 2021; Khatatbeh, Hani, and Abu-Alfoul 2020), we decided on limiting the time range of posts included in our study to ten days after lockdown imposition. Subsequently, based on SCCT and conceptual literature on COVID-19 (e.g. Bacq and Lumpkin (2021); Weaver (2020); Weaver and Blakey (2022)) the first author of this study drafted a coding scheme to identify compassion and MAVOON framings in the Facebook posts. Drawing a random sub-sample of 15 SEs from the total sample of 121 SEs, a first content analysis (Mayring 2004) was conducted by one independent coder. This way, we checked the suitability of our coding scheme and had the possibility to inductively add to the scheme based on the coding material. After a discussion of the independent coder and the first author of the study, there was a consensus that no addition to the coding scheme was necessary. Thus, the scheme was then applied to the full sample.

Based on the coding scheme, Facebook posts were coded as applying compassion framing if SEs reported (fear of) financial difficulties, a postponement or cancellation of product launches or events or difficulties to reach their customers/beneficiaries after the lockdown. Posts were coded as applying MAVOON framing if SEs saw opportunities to apply new technologies and forms of collaboration (e.g. video-conferencing) in the future, the potential to expand their range of customers (e.g. by offering online services), or the chance to apply their skills to address lockdown-related problems (e.g. by using innovative tools to support distance-learning for school students). We also created a category recording purely informative posts with no indication of SE-specific framings. This category included general information and advice related to COVID-19, for example, on how to behave during a lockdown (#stayathome).

The coding was done separately by one independent coder and the first author of this study. Comparing the results yielded a 96% inter-coder consensus. The remaining 4% were discussed and assembled to mutual satisfaction.

Quantitative analyses

Based on qualitative analyses, we employed a four-step procedure for subsequent quantitative analyses. First, we examined the robustness of our data. Doing so, we screened for common method bias, examined potential heteroscedasticity, and conducted an outlier analysis. Second, we used descriptive statistics to get an overview of the data, the share of COVID-19-related posts in general, and the distribution of posts applying compassion and MAVOON framing. Third, to answer RQ₁, we calculated two non-parametric Kruskal Wallis tests to investigate differences in compassion and MAVOON framing after the two lockdowns. This type of test was chosen as it is more robust than other mean difference tests like t-tests (Field 2018; Kruskal and Wallis 1952). Fourth, exploring RQ₂, we used linear mixed models to identify potential changes in (i) the total number of COVID-19-related posts, (ii) the number of posts applying compassion framing, and (iii) the number of posts applying MAVOON framing. Linear mixed models were applied, as they are more robust than ‘traditional’ repeated-measures analyses like analyses of variance. Linear mixed models allow to model non-linear relationships over time and can handle missing values better (Krueger and Tian 2004). Moreover, they do not require data-sphericity (Smith 2012). As control variables, enterprise age and the number of Facebook subscribers were included. Enterprise age could be important as young SEs are typically short of liquidity and personnel (Perrini, Vurro, and Costanzo 2010) and could find it particularly hard to deal with COVID-19-induced lockdowns. The number of Facebook subscribers was included to account for SEs’ visibility in the public. Previous research shows that, depending on media visibility, enterprise-stakeholder communication can be different (Ihmels et al. 2022).

Results

Robustness tests

We investigated the statistical robustness of our sample as follows. First, we checked for common method bias. This bias can occur if all data is retrieved from one single source (in our case, Facebook posts). Conducting Harman's single factor test by entering all data in a factor analysis and extracting one single factor yielded an amount of 20.58% variance explained. As this lies below the threshold of 50%, we consider common method bias not a serious problem in our sample (Fuller et al. 2016). Second, examining heteroscedasticity, we ran the White-test and the modified Breusch-Pagan-test separately for MAVOON and compassion framings after the two lockdowns (Field 2018). We found no significant results (Table 3). Thus, homoscedasticity can be assumed. Third, checking for outliers, we computed Mahalanobis distances for our data (Leys et al. 2018). Evaluating the results based on sample-size-adapted thresholds (Penny 1996), the data did not violate the assumption of equidistant scores. Consequently, no notable outliers featured in our data.

Table 3. Summary of White-test and modified Breusch-Pagan-test results on sample heteroscedasticity.

	Framing	White-Test	Modified Breusch-Pagan-Test
First lockdown (22nd March)	MAVOON	χ^2(5) = 2.37, p = .80	χ^2(1) = 0.95, p = .33
	Compassion	χ^2(5) = 1.18, p = .95	χ^2(1) = 0.31, p = .58
Second lockdown (16th December 2020)	MAVOON	χ^2(5) = 6.05, p = .30	χ^2(1) = 0.18, p = .67
	Compassion	χ^2(5) = 3.30, p = .66	χ^2(1) = 0.45, p = .50

Note. MAVOON: making-a-virtue-out-of-necessity; χ²(df): Chi-square value (degrees of freedom), values of p>.05 indicate homoscedasticity.

Data overview – descriptive statistics

As can be seen in Table 4, 299 Facebook posts were published up to ten days after the first lockdown. Of these posts, a total of 123 (41.14%) were related to COVID-19. 43.10% of all 123 COVID-19-related post featured MAVOON framings while 9.76% contained compassion framings. The remaining 47.15% were purely informative post on COVID-19 and featured no framings. Regarding the second lockdown, 360 Facebook posts were published up to ten days after imposition of which 30 (8.33%) were related to COVID-19. Of these 30 posts, 43.33% contained MAVOON framings while 16.67% featured compassion framings. The remaining 40.00% were purely informative and contained no framings. Interestingly, it turned out that, for both lockdowns, appx. 90% of COVID-19-related posts were posted in the first three days after lockdown imposition.

Table 4. Descriptive statistics of facebook postings (N = 121 social enterprises).

		M	SD	Sum
First lockdown (22nd March 2020)	All posts^a	2.471	3.225	299
	COVID-19-related posts^b	1.017	1.378	123
	MAVOON framing	0.438	0.752	53
	Compassion framing	0.099	0.300	12
Second lockdown (16th December 2020)	All posts^a	2.975	3.689	360
	COVID-19-related posts^b	0.248	0.505	30
	MAVOON framing	0.107	0.311	13
	Compassion framing	0.041	0.200	5

Note. ^aTotal number of posts up to ten days after lockdown imposition; ^bTotal number of COVID-19-related posts up to ten days after lockdown imposition; MAVOON: making-a-virtue-out-of-necessity.

Research question 1 – Kruskal Wallis test

To answer RQ₁ concerning the extent to which SEs applied MAVOON and compassion framings in their stakeholder communication after COVID-19-induced lockdowns, we used two non-parametric Kruskal Wallis tests. For lockdown 1, we found a significant coefficient (H_{Lockdown 1} = 21.07, p < .01). This indicates a difference comparing the frequency of both types of framing. Considering the higher number of MAVOON framings (cf. Table 4), a significantly higher share of MAVOON compared to compassion framings occurred. Results for lockdown 2 also yielded a significant difference (H_{Lockdown 2} = 3.85, p < .05). As the number of MAVOON framings is higher than the number of compassion framings after lockdown 2 (cf. Table 4), this indicates a significantly higher share of MAVOON framings. Based on Dunlap (1994), effect sizes of r = 0.41 for lockdown 1 and r = 0.15 for lockdown 2 emerged. Compared to the average psychological effect size of r = 0.21, this indicates an above-average effect size after lockdown 1 and a below-average effect size after lockdown 2 (Funder and Ozer 2020).

Research question 2 – linear mixed models

Examining potential changes in the application of MAVOON and compassion framing in SE-stakeholder communication comparing successive COVID-19-induced lockdowns, three linear mixed models were calculated. As can be seen in Tables 5–7, two significant and one insignificant coefficient occurred. We find a significant coefficient for the total number of COVID-19-related posts (I(112) = 0.82, p < .01). This indicates a significant difference (Table 5). As Table 4 yields, there is a significant decrease after lockdown 2. As can be seen in Table 6, we also found a significant difference for MAVOON framings (I(112) = 0.35, p < .01). This signifies that the number of posts containing MAVOON framings decreased after lockdown 2 (cf. Table 4). Finally, Table 7 shows that no significant coefficient emerged for compassion framings (I(112) = 0.04, p = .20). Thus, the number of posts containing compassion framings did not change comparing lockdown 1 and lockdown 2. In all three linear mixed models, our control variables had no significant coefficient.

Table 5. Summary of linear mixed model calculations for total number COVID-19 related posts.

Variable	Estimate	SE	df	t	p	95% CI
						LL	UL
Intercept	0.080	0.098	111.148	0.816	0.416	−0.114	0.274
IΔLockdown posts	0.823	0.130	112.000	6.322	0.000	0.565	1.081
Facebook subscribers	−2.922e^−7	3.993e^−7	110.000	−0.732	0.466	−1.084e^−6	4.992e^−7
Enterprise age	0.110	0.061	110.000	1.798	0.075	−0.011	0.231

Note. SE: standard error; df: degrees of freedom; CI: confidence interval; LL: lower level; UL: upper level; I: index variable for differences in number of posts comparing the first and second lockdown.

Table 6. Summary of linear mixed model calculations for MAVOON framings.

Variable	Estimate	SE	df	t	p	95% CI
						LL	UL
Intercept	0.052	0.056	117.312	0.932	0.353	−0.058	0.162
IΔLockdown posts	0.354	0.078	112.000	4.514	0.000	0.199	0.509
Facebook subscribers	−1.407e^−7	2.228e^−7	110.000	−0.631	0.529	−5.823e^−7	3.009e^−7
Enterprise age	0.034	0.034	110.000	0.993	0.323	−0.034	0.102

Note. MAVOON: making-a-virtue-out-of-necessity; SE: standard error; df: degrees of freedom; CI: confidence interval; LL: lower level; UL: upper level; I: Index variable for differences in number of posts comparing the first and second lockdown.

Table 7. Summary of linear mixed model calculations for compassion framings.

Variable	Estimate	SE	df	t	p	95% CI
						LL	UL
Intercept	0.023	0.034	132.265	0.681	0.497	−0.045	0.091
IΔLockdown posts	0.044	0.034	112.000	1.295	0.198	−0.023	0.112
Facebook subscribers	−6.456e^−8	1.319e^−7	110.000	−0.489	0.626	−3.260e^−7	1.969e^−7
Enterprise age	0.016	0.020	110.000	0.767	0.445	−0.025	0.056

Note. SE: standard error; df: degrees of freedom; CI: confidence interval; LL: lower level; UL: upper level; I: Index variable for differences in number of posts comparing the first and second lockdown.

Discussion

The current study combines qualitative and quantitative methodologies to explore SE-stakeholder communication after two successive COVID-19-induced lockdowns. Building on SCCT and framing literature, we use Facebook posts to examine the application of MAVOON and compassion framing in the SE context (RQ₁) and whether potential changes occur comparing two successive nationwide lockdowns (RQ₂).

Research question 1

Regarding RQ₁, our results suggest that SEs’ applied more MAVOON framings than compassion framings after both lockdowns. This seems plausible due to two reasons. First, in contrast to commercial enterprises, SEs’ primary target is social value creation (Kruse, Wach, and Wegge 2021). Considering the massive social problems caused by lockdowns like increased loneliness, marginalisation of certain groups or job losses (Schippers 2020), and the limited capacity of the state to ease lockdowns’ detrimental consequences, the number of potential beneficiaries rose (Ratten 2022; Weaver 2020). Thus, MAVOON framing can be seen as a convincing and credible framing strategy in SE-stakeholder communication. Second, while compassion framing is, generally, considered a strategy to strengthen the emotional bond between organisations and their stakeholders (cf. Coombs (1999)), in social entrepreneurship, it could turn out as a weakening of this bond. By excessively stressing financial problems stakeholders could get the impression of a shift in SEs’ interest from social value creation to money generation. This so called mission drift is a phenomenon that receives growing attention in scientific literature (Grimes, Williams, and Zhao 2019; Varendh-Mansson, Wry, and Szafarz 2020) and a (perceived) mission drift of SEs has been shown to be negatively associated with their stakeholder relationships (Klein, Schneider, and Spieth 2021). Thus, the relatively low rate of compassion framing applied by SEs could be linked to the intention to avoid the impression of a mission drift.

Research question 2

Analysing RQ₂ on potential changes in the application of MAVOON and compassion framings comparing two successive lockdowns, yielded different results. On the one hand, the number of MAVOON framings decreased after the second lockdown. The reason for this could be two-fold. First, some chances identified after the first lockdown could have been realised until the second lockdown. Thus, they could not be eligible for the future-oriented MAVOON framing any more. To exemplify, digital collaboration tools like video-conferencing had to be installed quickly to ensure communication under social-distancing conditions (Härting et al. 2022; Pinzaru, Zbuchea, and Anghel 2020). Thus, when the second lockdown started, they were no longer eligible as a ‘new opportunity’ in a MAVOON framing. Second, some chances communicated after the first lockdown could have turned out as unrealistic. To illustrate, while a potential of serving more beneficiaries exists, the extension of an operational range with no direct access to potential beneficiaries is very challenging (cf. Shah et al. (2020)). Thus, chances perceived after the first lockdown could have originated from a too simplistic perception of a complex reality and, as a result, had to be dropped. Previous work on the prevalence of biases in social entrepreneurship supports this assumption (Chipeta, Kruse, and Venter (2022); Chipeta, Venter, and Kruse (2022); Gras, Conger, and Cummings (2017), Starr (2016), for examples).

On the other hand, we found no significant change in the number of compassion framings. Consequently, the framing ratio of MAVOON and compassion framings tends more towards compassion framing in the second compared to the first lockdown. In addition to the abovementioned potential factors contributing to a decrease in MAVOON framings, the consistent share of compassion framings could be associated with a similar (or even more severe) financial threat perceived by SEs after the second lockdown.

Notably, we also find that the total number of COVID-19 related posts decreased after the second lockdown. This could be linked to the relative ‘familiarity’ with the topic. As information novelty is essential for (social) enterprises to attract Facebook users (Tafesse 2015), posts only ‘refreshing’ already-known COVID-19 and lockdown-related information could have been omitted.

Implications for research

The current study’s implications for research include the following:

First, our study is one of the first empirically investigating how SEs communicate with their stakeholders. Thereby, we overcome the wide-spread ‘heroic bias’ (cf. Brattström and Wennberg (2022) in social entrepreneurship and emphasise the importance of stakeholders and adequate stakeholder communication. This complements previous studies. Furthermore, as suggested by several scholars, our work acknowledges the relevance of stakeholders for SE success (Dorado et al. 2022; Kruse, Chipeta, and Ueberschär 2023; Saebi, Foss, and Linder 2019; Van Wijk et al. 2019). Thus we account for the embeddedness of SE actions in a bigger context which is essential to paint a more realistic picture of the complex SE phenomenon.

Second, we extend previous literature linking COVID-19-induced lockdowns and SEs by examining and comparing two successive lockdowns. As our data showed, notable changes in stakeholder communication occurred. Future studies should dive deeper into these changes and try to identify their reasons and underlying mechanisms.

Third, applying a mixed method approach, our study follows the recommendation by Gupta et al. (2020) to combine qualitative and quantitative methodologies. Despite notable exceptions (e.g. Kruse (2021)), mixed method approaches remain underrepresented in SE research despite their great potential to gain more comprehensive insights compared to the application of one methodology alone. We hope that our work makes more scholars aware of the potential benefits of mixed method approaches. This could help to advance social entrepreneurship research further.

Fourth, analysing Facebook data, we use a largely untapped data source in social entrepreneurship. While we acknowledge the limitations of social media data (cf. section 5.5) we concur with Kaiser and Kuckertz (2023) that this data source can complement ‘traditional’ scientific data retrieved from interviews or questionnaires, particularly when examining enterprise communication.

Implications for practice

Our study yields the following practical implications:

First, SE-scholars can benefit from our findings when planning similar studies in the future. While we chose to code all COVID-19-related posts up to ten days after lockdown announcement, it turned out that appx. 90% of Facebook posts occurred in the first three days after lockdowns had been announced. Thus, considering that the coding interval of ten days was arbitrary, we recommend shorter coding intervals of three days. This way, similar research can become more efficient as the number of posts to be screened reduces notably while information loss is relatively small.

Second, through a managerial lens, our results yield that cognitive framings are applied in SE crisis communication. While, by definition, crises encompass a loss of control and pose a serious threat to enterprise survival, framings can serve as a tool for managers to actively shape stakeholder communication and influence stakeholder perception of how the crisis is managed.

Third, for policymakers, our finding that the MAVOON/compassion-framing ratio worsened after the second lockdown can provide notable information. While the reasons for communicating less chances after the second lockdown are beyond the scope of our study, it could be an indication of the need for long-term and sustainable government support in times of crisis. Despite large-scale financial and administrative support in initial stages of nationwide lockdowns (cf. Thorgren and Williams (2020)), the extent of support decreased in many countries. However, considering large governmental spending on measures easing social pressure caused by lockdowns (or other crises), funding for SEs could be an investment and help governments to alleviate social problems in a sustainable way (Kuckertz et al. 2023; Serwanga et al. 2014; Shomoye-Olusi et al. 2022).

Fourth, as our study illustrates how SEs apply cognitive framings in crisis communication, we argue that crisis communication should receive more attention in SE-courses. Considering the high density of global and local crises in the last 20 years (Aldao et al. 2021), competencies in effective crisis management gain importance in social entrepreneurship. Thus, insights on how cognitive framings can be used to actively shape stakeholder communication in a crisis can be helpful for nascent social entrepreneurs. SE-educators could use SCCT (Coombs 2007) as a basis to design corresponding educational modules.

Limitations

The current study’s limitations are as follows:

First, only German SEs with a SEND-membership were investigated. This limits our findings’ generalisability as the sample drawn is not representative for the whole German SE-landscape. Furthermore, SEs in our sample operated only in a limited number of sectors and their geographical distribution was uneven.

Second, as no (matched) control group of commercial enterprises featured in our study, it is not possible to compare the application of MAVOON and compassion framings. Thus, it remains unclear whether notable differences between framings in SEs and commercial enterprises exist. In addition, as no experimental design was used in our study, we cannot draw causal conclusions (Kruse 2020b).

Third, in response to the COVID-19 pandemic, countries took several different measures. Despite using the umbrella term ‘lockdown’, notable inter-country differences can be remarked (cf. Koh (2020)). Thus, our study’s findings cannot be transferred to countries with other lockdown measures and their robustness in other contexts needs to be checked.

Fourth, while very similar measures were taken during the first two nationwide lockdowns in Germany, infection rates were not equally distributed. Different infection rates could have affected/biased SEs’ communicational framings in areas with higher or lower rates.

Fifth, our study was restricted to COVID-19-induced lockdowns as one specific victim crisis. As Coombs and Holladay (2015) outline, notable other forms of crises exist that might result in different communicational framings when interacting with stakeholders. Consequently, our results are not transferrable to other types of crises like accidents.

Sixth, despite notable advantages when using social media data, there is the substantial risk of retrieving biased information. Positivity bias, i.e. a selective and overly positive reporting of events (Schreurs and Vandenbosch 2021) is a serious problem on social media and could have biased our results in favour of MAVOON framings.

Summary

As a response to the COVID-19 pandemic, lockdowns, i.e. restrictive government measures encompassing social distancing and the closure of almost all offices, public facilities, and shops, were imposed to contain SARS-CoV-2 infections. While first insights suggest that social enterprises, i.e. hybrid enterprises applying entrepreneurial means to fulfil their social missions, have chances and risks facing COVID-19-induced lockdowns, research remains surprisingly silent on how social enterprises communicated with their stakeholders. Our study addresses this gap and examines the application of two different stakeholder communication styles (MAVOON and compassion framing) by 121 German social enterprises after two successive lockdowns. Building on a mixed methods approach and analysing Facebook data, we can show that chance-oriented MAVOON framings are more frequently applied than risk-oriented compassion framings after both lockdowns. However, the number of MAVOON framings decreased after the second lockdown. Despite suffering from a limited generalisability and a potential positivity bias, our study makes original contributions to social entrepreneurship research and practice in three ways. First, our work sheds light on SE stakeholder communication processes after COVID-19-induced lockdowns and uncovers changes in SEs’ communicational framings. Second, we examine real SE-stakeholder communication using social media data as a largely untapped data source in SE-research. Our study makes recommendations for similar research in the future, for example, regarding coding intervals. Third, practitioners can benefit from our study in different ways. We highlight cognitive framings as a tool for SE-managers, emphasise the importance of sustained and long-term policymaker support for SEs in times of crises, and encourage SE-educators to consider crisis communication as an important element in SE-courses.

Acknowledgements

I would like to thank Ms. Nicolette Heinze for her invaluable support in data acquisition and handling.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data used in this study is available upon reasonable request to the first author.