Emotional framing of NGO press releases: Reformative versus radical NGOs

Drawing on the literature on framing, we explore the emotional framing differences in radical and reformative NGOs over time. We analyse the sentiment of a sample of 5880 press releases issued by five NGOs positioned differently on the reformative-radical spectrum and examine how they address large companies. Our findings reveal an increasing polarisation of sentiment in these NGOs' framing, with individual NGOs gravitating towards ideal-type radical or reformative positions, respectively. In alignment with the differences in their framing, we observe differences in their approaches to cross-sector partnerships. Policymakers need to note the implications of the observed polarisation for the effectiveness and credibility of cross-sector partnerships and multi-stakeholder initiatives more generally, given the risk of co-optation (for reformative NGOs) as well as the risk of foregoing significant funding and governance opportunities (for radical NGOs). their optimism towards businesses, us their positioning radical-reformative spectrum. We content illustrate further the role of cross-sector partnerships in the NGOs' framing.


| INTRODUCTION
Non-governmental organisations (NGOs) have emerged as relevant policy actors at various levels due to their role in motivating corporate action on various environmental and social issues (Yaziji & Doh, 2009). How they motivate corporate action ranges from smear campaigns that use an adversarial language and antagonistic strategies to more constructive cross-sector partnerships where NGOs use a friendly language and act as external consultants or change agents (Burchell & Cook, 2013a). Indeed, these differences in NGO-business relationships reflect the reformative-radical spectrum that helps us categorise NGOs.
Reformative (or insider) NGOs try to improve and reinforce existing institutional structures, whereas radical (or outsider) NGOs explicitly try to change or undermine existing institutional structures (den Hond & de Bakker, 2007;Yaziji & Doh, 2009;Young, 1999). The reformative versus radical categorisation is not a static one: instead, it is a spectrum whereby NGOs may shift their positions over time. Such shifts on the reformative-radical spectrum are apparent in the language NGOs use to address businesses. Hence, by analysing the changes in their language, we can capture how NGO positions change over time. Exploring such change is critical, because these changes can reveal their framing (Litrico & David, 2017), which is also associated with their attitude towards cross-sector partnerships (Klitsie, Ansari, & Volberda, 2018). While a recent study demonstrates shifts in framing in a single longitudinal case (Luxon, 2019), radical and reformative NGOs' framing would differ significantly and be likely to show different evolutionary patterns.
In this paper, our objective is to demonstrate these shifts in radical and reformative NGOs' framing in a longitudinal cross-case research design and explain cross-sector partnerships' role in these shifts. To do so, we draw on the literature on framing (Bach & Blake, 2015;Baek & Yoon, 2017;Benford, 1993;Benford & Snow, 2000) and the literature on cross-sector partnerships and NGO-business relationships (Baur & Schmitz, 2012;Bitzer & Glasbergen, 2015;den Hond, de Bakker, & Doh, 2015;Pesqueira, Glasbergen, & Leroy, 2020). Given the prominence of major international NGOs impacting corporate environmental action, we select the following five cases-Conservation International, Friends of the Earth, Global Witness, Greenpeace and WWF-between 2002 and 2017.
We adopt the well-established view that actors motivate action by giving sense to their stakeholders through various sentiments, also referred to as emotional framing (Pesqueira et al., 2020;Raffaelli, Glynn, & Tushman, 2019;Reber & Berger, 2005;Reinecke & Ansari, 2016). It is common to study framing shifts in organisational communication of NGOs, including their social media accounts (Ji, Harlow, Cui, & Wang, 2018), news media coverage (Brulle & Benford, 2012) and press releases (Luxon, 2019). We also study 5880 press releases issued by these NGOs and investigate framing shifts concerning 225 of the largest multinational companies (MNCs). We analyse these press releases using a dictionary-based sentiment analysis tool and examine their degree of optimism towards businesses, which helps us understand their positioning on the radical-reformative spectrum. We also conduct content analysis to illustrate further the role of cross-sector partnerships in the NGOs' framing.
Our results reveal framing shifts that reflect an increasing polarisation between radical and reformative NGOs. We find that over time, NGOs move further towards ideal-type radical or reformative positions, respectively. Crucially, those NGOs that have been explicitly open to various types of cross-sector partnerships with MNCs have moved towards more reformative positions. In contrast, those continuing to reject such partnerships have moved towards a more radical stance over time. This polarisation has implications for the role of these NGOs in wider governance. At one end of the continuum, it increases the likelihood of NGO co-optation, in turn carrying significant risks for the credibility and legitimacy of NGOs that adopt a collaborative stance (Baur & Schmitz, 2012). At the other end of the continuum, given the dominance of the partnership paradigm in global governance (Pearce & Doh, 2005), there is a risk that more radical NGOs are becoming increasingly excluded from private governance processes.
To explore these dynamics, our paper provides a longitudinal cross-case analysis of NGOs positioned differently on the reformative-radical spectrum. Since the literature has mostly been dominated by conceptual work and case-based inquiries in this area (den Hond, 2010;den Hond & de Bakker, 2007;Yaziji & Doh, 2009), our approach contributes to the literature in three ways. First, we empirically identify trajectories of framing shifts at radical and reformative NGOs and show these shifts using sentiment analysis. By doing so, we add to the existing conversation on framing in the context of NGOs (Luxon, 2019;Pesqueira et al., 2020) and NGO-business relationships (Klitsie et al., 2018;Le Ber & Branzei, 2011;Zimmermann, 2020). Second, we demonstrate that NGOs' emotional framing varies depending on their position in the radical-reformative spectrum (Luxon, 2019). Third, the application of sentiment analysis to study the framing of NGO press releases is a promising methodological contribution to the broader literature on NGO-business relationships. Measuring the sentiment of NGO press releases allows us to capture changes in actual NGO campaigning directly, whereas previous empirical studies have focused on survey responses (den Hond et al., 2015) or variables such as board interlocks and annual report content (Bertels, Hoffman, & DeJordy, 2014). Only recently, sentiment analysis has been applied in the context of NGOs to capture the communication strategies of NGOs (Luxon, 2019) or media frames about NGOs (Wasif, 2020). Still, these efforts have not explicitly explored emotional framing in the context of NGO-business campaigning.
The remainder of this paper is structured as follows. The next section reviews the previous literature on NGOs and framing, with particular emphasis on NGO-business relationships. The methods section describes and justifies case selection, data collection and data analysis. Next, we present the results of the empirical analysis. We then discuss the implications of these results for theory, practice and policy. We conclude by spelling out the limitations of our approach as well as promising avenues for future research.

| NGO definitions and categorisation
NGOs are 'private, not-for-profit organisations that aim to serve particular societal interests by focusing advocacy and/or operational efforts on social, political and economic goals, including equity, education, health, environmental protection, and human rights' (Yaziji & Doh, 2009, p. 5). In its early stages, the NGO scene could be described as 'anarchic' (Winston, 2002) and comprised a heterogeneous group of organisations. Over time, however, we observe an increasing consolidation and diversification, with different NGO types pursuing fundamentally different strategies to inform and mobilise the general public around environmental and social challenges (Brown, Khagram, Moore, & Frumkin, 2000). Table 1 below provides an overview of contemporary NGO typologies.
At a very general level, NGOs can be distinguished based on the governance functions they aim to fulfil. Service sector NGOs try to meet previously unmet needs with goods and services they provide, whereas advocacy NGOs aim to bring about institutional change through policies or practices (Yaziji & Doh, 2009). Among advocacy NGOs, two main types can be identified. Reformative NGOs work within the existing institutional structures to bring about incremental change; they may even 'reinforce the institutional environment by challenging infringements on it' (den Hond, 2010, p. 175). Radical NGOs aim directly at changing the institutional environment (Brown, 2010;den Hond, 2010;den Hond & de Bakker, 2007;Laasonen, Fougère, & Kourula, 2012), with some authors identifying additional configurations that are positioned between these two extremes (Ählström & Sjöström, 2005;van Tulder & van der Zwart, 2005). The distinction between reformative and radical NGOs then captures two fundamentally different NGO-business relationships (Fisher, 1997). Depending on their positioning on the reformative-radical spectrum, NGOs apply different means to reach their objectives (den Hond & de Bakker, 2007;Winston, 2002;Yaziji & Doh, 2009). These differences are, indeed, associated with the emotional framing they adopt in their campaigning.

| Framing
Goffman (1974, p. 21), who popularised frames and frame analysis in sociology, refers to frames as 'schemata of interpretation' that 'allow [their] users to locate, perceive, identify and label a seemingly infinite number of concrete occurrences.' Picked up by scholars of social movements, the concept has frequently been used as a dynamic process of 'framing' 'that implies agency and contention' (Benford & Snow, 2000, p. 614). These scholars specifically study what they call 'collective action frames' as the outcomes of framing processes that are 'action-oriented sets of beliefs and meanings that inspire and legitimate the activities and campaigns' (Benford & Snow, 2000, p. 614). However, the application of framing has been quite broad, spanning from politics (Coutto, 2020;Iyengar, 1990) to media and communication studies (Collins, 2008;Du & Han, 2020;Engesser & Bruggemann, 2016;Entman, 1991) as well as business and organisation studies (George, Chattopadhyay, Sitkin, & Barden, 2006;Girschik, 2018;Hoffman & Ventresca, 1999;Howard-Grenville & Hoffman, 2003). Two distinct approaches in these framing studies can be identified: studies that focus on frames as relatively stable constructs versus those that approach framing as a dynamic interactional process, focusing on shifts in framing (Cornelissen & Werner, 2014;Gray, Purdy, & Ansari, 2015;Zimmermann, 2020). In this study, we adopt the latter perspective.
As a process, framing connotes actors' efforts to provide meaning for others, that is, to influence their interpretations as a form of sense giving to change their actions (Fiss & Hirsch, 2005;Fiss & Zajac, 2006;Gray et al., 2015). There are many different attributes of framing that scholars have researched. Some scholars have researched how actors construct specific meanings around environmental and societal issues, also referred to as issue framing (Andersson & Bateman, 2000;Bach & Blake, 2015;Dewulf & Bouwen, 2012). Others have researched the role of 'core framing tasks' (Snow, Vliegenthart, & Ketelaars, 2018): to shape the past (diagnostic) of a problem, how its future (prognostic) should be and giving meaning to others as a call to action (motivational) (Hervieux & Voltan, 2016;Hestres, 2018). Due to the dominant role of rationality in the broader social science research, the least researched attributes of framing have traditionally been emotions and sentiments. However, this has started to change lately with the growing interest in 'emotional framing' (Eyerman, 2005;Raffaelli et al., 2019;Ruiz-Junco, 2013;Snow et al., 2018). This is also the area in which we position our study.

| Framing of NGO-business relationships
Emotion and sentiment are especially critical attributes of framing in the context of NGOs. NGOs use 'an emotionally-laden narrative' to create urgency and to alarm stakeholders, triggering their feelings to mobilise action (Reinecke & Ansari, 2016). They set an emotional tone when engaging with their stakeholders to alter their perceptions using negative messages by creating feelings of fear, anger, guilt, shame or sadness (Anspach & Dragulji c, 2019;Gifford & Comeau, 2011). They may also frame issues with positive messages by evoking hope and faith, and implying benefits and opportunities as well as signalling connectedness with their cause through love and joy (Flam & King, 2007). The emotional attributes of NGOs' framing are not independent of whether the NGO is radical or reformative (Luxon, 2019), and whether it targets the 'enemy' or a 'friend' (Knight & Greenberg, 2011). In this paper, we specifically focus on NGObusiness relationships that may include both antagonistic and co-operative relationships and cross-sector partnerships between NGOs and businesses as a specific form of co-operative relationships.
Reformative and radical NGOs differ in the extent to which they integrate market, community and environmental logics (Lee & Lounsbury, 2015;Lounsbury, Ventresca, & Hirsch, 2003), affecting the messages they give to their stakeholders about businesses, hence impacting the emotional tone of their framing. Reformative NGOs mainly represent functional units with little political purpose, utilising their specific expertise to address social problems (Ossewaarde, Nijhof, & Heyse, 2008). They are often 'regarded as part of marketbased solutions to policy problems' (Lewis & Kanji, 2009, p. 5). Therefore, one would assume their framing to be more positive than their radical counterparts. In contrast, radical NGOs aim to act as 'a countervailing power to the market and the state, adding to the creation of an international system of checks and balances' (Ossewaarde et al., 2008, p. 43), thus explicitly positioning themselves as political actors. In alignment with this position, one would assume that radical NGOs apply an adversarial framing to trigger shame and guilt in the businesses they target and create anger and distrust among their followers (Flam & King, 2007;Knight & Greenberg, 2011); which can generally mean a more negative tone when compared to their reformative counterparts.
However, it is unlikely to expect stability in the framing efforts of NGOs. External events such as climate summits (Ansari, Wijen, & Gray, 2013), exogenous shocks such as financial crises (Kamat, 2004) and relationships NGOs have with global policy institutions and donors (Kamat, 2003) may change and evolve in different directions over time. Here, our objective is to explore and explain how they evolve, which would be observed as shifts in their framing, and therefore changes in their emotional tone or sentiment. Based on the existing literature, it is possible to put forward arguments for a shift towards both a more reformative and a radical stance. There may be several explanations as to why NGOs may shift from a radical position to a more reformative one. Studies propose that after the initial stages of NGO advocacy that were characterised as mostly antagonistic, NGOs have evolved in the direction of increasing involvement and engagement with MNCs (Doh & Guay, 2004;Heap, 2000;Valor & Merino de Diego, 2009). This is because reformative positions typically offer benefits such as improved access to funding and increased participation in global level decision-making (Arya & Salk, 2006;Pearce & Doh, 2005). partnerships have come to be viewed as more effective than adversarial approaches (Pearce & Doh, 2005). Furthermore, these partnerships may lead to the adoption of 'business-like practices, such as the use of formal organisational charts, quantitative evaluation accounting systems, and independent financial audits' at NGOs (Ahmadsimab & Chowdhury, 2019, p. 14). Hence, such a shift would imply a change in NGOs' emotional tones toward businesses from a more negative into a more positive direction.
Equally, there may be several explanations as to why NGOs may shift from a reformative to a more radical position. At a practical level, failed projects may lead to 'partner disillusionment' (Le Ber & Branzei, 2010 in NGOs losing sight of their initial ideologies and thus lead to a deradicalisation of these organisations (Baur & Schmitz, 2012;Ossewaarde et al., 2008). Some criticised the adoption of businesslike practices that hint the NGOs' reformative stance (Sanders & McClellan, 2014); while others problematized the effectiveness of cross-sector partnerships (Laasonen et al., 2012). As a consequence, NGOs may shift towards a more radical stance, which would imply a change in NGOs' emotional tones toward businesses in a more negative direction.

| Multiple case study: selection of NGO cases
We based our case selection on the following criteria: (1) they operate globally and are therefore participating in global-level environmental governance; (2) they have self-published public archives of press releases that both cover the duration of the analysis and are sufficient in number to establish the statistical significance of any findings; and (3) reflect a spectrum ranging from radical to reformative NGOs. We selected the following cases: Conservation International, the World Wildlife Fund, Friends of the Earth, Global Witness and Greenpeace.
Among the NGOs that have traditionally leaned towards the spectrum's reformative end, Conservation International (CI) was established in 1987 in the United States. In its initial stages, it had a pure conservation focus; recently, CI has started to take a wider perspective and explicitly focus on the link between human well-being and natural ecosystems (CI, 2021).
As another reformative NGO, the World Wildlife Fund for Nature WWF is actively engaged in partnerships with the private sector, from which it receives a significant amount of funding and can, therefore, be characterised as reformative (Ählström & Sjöström, 2005;den Hond, 2010).

| Data collection
The media's importance as an arena in which NGOs, or social actors more generally, construct norms or legitimacy is highly significant (King & Soule, 2007;Richards & Heard, 2005;Siltaoja & Vehkaperä, 2010 In line with this stream of research, we collected the five NGOs' official press releases. These press releases are designed to provide meanings and give sense to other stakeholders (Fiss & Zajac, 2006) and, therefore, are well-suited for the analysis of emotional framing adopted by NGOs (Luxon, 2019). Moreover, these press releases can impact public legitimacy and thus the reputation of a range of companies (Pearce & Doh, 2005); to a certain extent, they equal social action in this context. The extensive database of press releases we employed, spanning from 2002 to 2017, allowed us to take a longitudinal perspective and explore framing shifts over time. Table 2 summarises our data collection. Our timeframe ranges from 2002 to T A B L E 2 Sample employed for sentiment analysis text press releases on their websites.

| Data analysis
3.3.1 | Sentiment analysis Sentiment analysis presents a great potential to study the emotional framing of NGOs in their communications through various types of media (Wasif, 2020), including social media (Åkerlund, 2020;Ji et al., 2018) and press releases (Luxon, 2019). While sentiment analysis has frequently been used to study framing in communication studies (Young & Soroka, 2012), its applications to business and organisation studies are more recent (Isil & Hernke, 2017) and yet to be explored in the context of NGO-business relationships.
For the purposes of this study, we measured the emotional tone of official NGO press releases and calculated their 'degree of optimism', which provides us with insights about the NGOs' framing (Luxon, 2019;Stieglitz & Dang-Xuan, 2013). Press releases expressing optimism towards business are generally positive and non-critical and, therefore, can be considered to represent a more co-operative stance, thus mirroring a reformative perspective on NGO-business relationships (den Hond & de Bakker, 2007;Yaziji & Doh, 2009;Young, 1999). In contrast, press releases expressing pessimism are generally more critical and represent a more adversarial stance, aligning with a radical position on NGO-business relationships. (Brown, 2010;den Hond, 2010;den Hond & de Bakker, 2007;Laasonen et al., 2012). Hence, we assumed that the degree of optimism could reveal NGOs' framing and, thus, their positioning on the reformative-radical spectrum.
We conducted a sentiment analysis of the sample of press releases spanning the years 2002-2017 (n = 5880; see Table 2), using the WordStat software package. We used the Forbes 2000 list of the largest companies for the year 2010 (as the median value in our period of analysis) and identified the 15 largest companies within 15 different sectors, therefore arriving at a set of 225 companies. All results were checked manually to ensure that these press releases referred to a given company (for example, to distinguish press releases mentioning the company 'Apple' from those using the term 'apple' in other contexts). All results were transcribed into an SPSS database for subsequent statistical analysis.
The basic underlying assumption of sentiment analysis is that the sentiment of a body of text-such as its optimism-can be revealed by the frequency of words of a certain type used by the writer. These types have been provided by the existing psycho-social dictionaries included in the WordStat software package. The WordStat Sentiment Dictionary is partially based on several individual sentiment lists, including the 'Harvard IV' dictionary as well as the 'Linguistic and Word Count' dictionary (Tausczik & Pennebaker, 2010). It is a general-purpose dictionary applied in various contexts, including corporate sustainability reporting (Lock & Seele, 2016), academic articles (Mora, Wu, & Panori, 2020) or policy documents (van Alstine & Barkemeyer, 2014). It provides individual scores for positive and negative sentiment, each calculated based on terms comprising several thousand-word patterns, respectively, and is therefore well-suited for our analysis. Appendix A illustrates the use of the dictionary-based sentiment analysis tool, demonstrating the coding of two press releases with very high or very low optimism scores, respectively.
Both positivity and negativity scores were expressed as the number of words indicating positive (negative) sentiment divided by the total number of words in a given press release to account for variable document length. We calculated the raw sentiment score for each press release by subtracting its negative from its positive score. Given that different NGOs cannot employ the same language, we transformed all raw sentiment data into normalised z-scores. First, averages and standard deviations were calculated for each of the five NGO subsamples. The difference between the NGO-specific mean and the raw score of a given press release was divided by the standard deviation of the NGO-specific mean.
Both positive and negative sentiment dictionaries were screened manually to ensure that the underlying terms were suitable for our analysis and to remove domain-specific words (cf. Loughran & McDonald, 2011). Furthermore, McKenny, Aguinis, Short, and Anglin (2016) highlight the risk of potential bias in an automated content analysis based on what they call algorithm error. In short, different software packages use different algorithms to assess positive and negative sentiment, leading to different results. For this reason, we replicated our analysis using the General Inquirer (GI) software package (Stone, Dunphy, Smith, & Ogilvie, 1966). Like WordStat, GI is a dictionary-based sentiment analysis tool based on extensive terms lists that indicate positive or negative sentiment; like WordStat, it also provides separate positivity and negativity scores. All results based on GI sentiment scores were qualitatively equivalent to the WordStat-based regression models reported in this study, thereby minimising the risk of bias due to algorithm error.

| Qualitative content analysis
We conducted a qualitative content analysis of press release content to complement the sentiment analysis, using NVivo 12 Plus. To further illustrate the cross-sector collaboration patterns between the NGOs we selected, we first searched the database based on terms associated with partnerships (Appendix B for the full list of search terms). We used the search terms that previous review articles on partnerships have identified (Dzhengiz, 2020;Gomes, Barnes, & Mahmood, 2016;Zahoor & Al-Tabbaa, 2020). We further complemented these search terms with synonyms of partnerships, alliances and collaboration in the Thesaurus. This helped us identify and qualitatively code the cross-sector partnerships between the NGOs and other organisations. This coding provided partnerships between these NGOs and businesses and other organisations, such as various other NGOs, universities and public authorities. Organisations other than businesses were beyond the scope for this study and therefore not considered, since we are only interested in NGO-business relationships. We, then, manually screened the press releases that contained these partnership-related terms and coded the NGOs and their business partners.

| Sentiment analysis
In this section, we report our findings, including the descriptive statistics regarding the industry and company coverage of NGO press releases, the degree of optimism reflected by the five NGOs combined, for reformative and radical NGO constituents individually, and their different approaches to cross-sector partnerships with MNCs. Table 3 provides a set of descriptive sample statistics as well as the results of the sentiment analysis. We observe that press releases are not spread evenly across the five NGOs. GP (n = 2068) and WWF (n = 1423) have been more active than the other three NGOs within the period under analysis. Furthermore, the NGOs differ in the extent to which they addressed any of our 225 focal companies (see Appendix C). FOE and GP have a stronger corporate focus than their peers with 29% and 25%, respectively, of their press releases mentioning at least one out of these 225 companies, as opposed to 8-18% in the case of CI, GW and WWF. Overall, 1109 press releases mentioned any of the focal companies, amounting to 19% of the total sample. Mean optimism z-scores (M = −0.125) of business-related press releases appear to be largely in line with the overall sample. However, clearer deviations can be identified at the level of individual NGOs, with average optimism scores ranging from −0.294 (FOE) to 0.398 (CI).
It is interesting to note that different NGOs tend to prioritise companies from different sectors. GW has a strong banking and oil and gas focus, with 145 out of 177 press releases addressing companies from these two sectors. The mining sector is more closely associated with FOE, whereas chemical companies are more likely to be scrutinised by GP. Oil and gas (n = 444) emerges as the sector that has received the most attention across the five NGOs. At the other end of the spectrum, hardly any coverage of companies in the construction, pharmaceutical or transportation sectors can be identified in these years. Table 4 provides an overview of the 15 most frequently mentioned companies in the sample (n ≥ 35). Here, chemical and in particular oil and gas companies again clearly receive the most attention. Three electronics companies are represented in the list of most frequently mentioned companies, even though electronics featured less prominently in the sector-level comparison in Table 3. Here, press releases related to two major campaigns-the GP 'Greener Electronics Guide' and the WWF 'Climate Savers' programme-typically list a range of electronics companies. Furthermore, patterns emerge in terms of specific companies that are prioritised by different NGOs.
For example, Dow Chemical has frequently been targeted by GP but not featured by the four other NGOs; FOE and GP have almost exclusively targeted Monsanto in the context of genetically modified organisms.

| Qualitative content analysis
The press releases also showed stark differences between NGOs regarding cross-sector partnerships with MNCs (see Table 5  GP also engaged with corporations; however, its engagement is less direct than WWF and CI. Instead, GP's corporate engagement appeared as embedded into campaigns such as its 'Detox Campaign', 'Greener Electronics Guide' and the 'Consumer Goods Forum'. This indirect involvement allowed GP freedom to report on companies' progress and take a critical stance if necessary. For instance, Adidas has been scrutinised by GP even after the company's commitment to GP's detox campaign and only after setting a credible roadmap has GP announced these commitments as a victory for the industry. FOE and GW did not report any partnerships with the business world and even outright rejected the effectiveness of cross-sector partnerships with MNCs as a mechanism to address environmental problems.
Some of the cross-sector partnerships announced turned controversial. In the case of BP and CI, this was due to the Deepwater Horizon oil spill. CI chairperson commented on the tragedy as follows: 'In 2001, CI began to work with BP on various initiatives to reduce their environmental footprint [which] included the Energy & Biodiversity Initiative, a collaboration with several oil and gas F I G U R E 2 Sentiment scores of business-related press releases over time per NGO (2002NGO ( -2017 T A B L E 5 NGOs and their approach to cross-sector partnerships and market-led initiatives 'We absolutely must work together to find new ideas and solutions to the daunting challenges before us. Therefore, we will always come to the table to have conversations --with communities, with governments, with scientists and with businesses that express an interest in doing their part to help us tackle these challenges. Failure to do so would be irresponsible. What was recorded on the tape and emails was simply an initial conversation with a corporation that approached us. But initial conversations are only a first step in a long process of due diligence about an organisation's commitment to environmental leadership. We challenge and collaborate with companies to improve their business practices and invest in conservation initiatives. Our commitment is to engage with corporations to minimise environmental impact and encourage them to proactively participate in programs to preserve healthy ecosystems and biodiversity; this is in their enlightened self-interest …' '… As the incubator for this business, Conservation International Hawai'i will provide the technical expertise and start-up funding and infrastructure for the successful establishment of the business. Profits from the business will be directed toward ocean stewardship and sustainable fishery initiatives…' WWF Coca Cola, HSBC, Toyota, Ikea, Netflix '… The Climate Savers companies show that sustainable development is not an academic concept but something that can be tackled with a profitfor nature, for society, but also for the companies themselves …' 'We know the commitments under the Heart of Borneo Declaration cannot be achieved without the support of the private sector. Tonight, we highlight solutions for involving business in green growth and offer then a range of tools to help them to do that.' '… Samantha Smith, leader of WWF's Global Climate and Energy Initiative says cleantech is finally coming of age. "Since the first Global Cleantech Innovation Index was produced just two years ago, we have seen a real increase in the disruptive start-ups we need to shrink our footprint on this planet. And that increase is spread across different countries, with some leading but all making progress. Of course, much more needs to be done in every country and by investors if we are to properly address climate change and achieve a transition towards a 100% renewable energy future …"' '… WWF was a co-founder of the MSC and openly promotes MSC certified products. WWF considers that the full application of the MSC's rigorous and robust standards and procedures is critical to ensuring the maintenance of its leadership position ….'

GP Corporate partners within 'Detox
Campaign', 'Greener Electronics Guide' and the 'Consumer Goods Forum' and Facebook '… Although Greenpeace is best known as a fierce opponent to corporate polluters, the organisation also works in cooperation with big business when corporate leadership is ready to transform its actions on behalf of the environment …' '… This case shows that the FSC needs to urgently establish certification safeguards in high risk areas where there are high levels of corruption and where good governance, the rule of law and organised civil society are all lacking …' '… Greenpeace calls on other palm oil producers to follow GAR's initiative rather than hide behind weak legislation and certification systems, such as the Roundtable on Sustainable Palm Oil (RSPO). If the palm oil industry is to erase the bad reputation it has acquired among consumers and financial institutions, it needs to be clear which plantation companies are truly committed to protecting Indonesia's remaining forests, and those which will continue their destructive practices …'

FOE None
'The UN has been working very closely with big business in developing and promoting the concept of "Green Economy" which is selling out nature and people and greenwashing a broken and unfair economic system at the expense of sustainable development …' 'As an alternative to the ICC awards, Friends of the Earth is part of the Greenwash Academy which itself is presenting awards to corporations which are greenwashing their bad social and environmental practices. FoEI has exposed the real story of big business bad practices around the World in its 'Attempts to use certification schemes to reduce the widespread environmental and social problems caused by growing crops for fuels and animal feeds are bound to fail, states a new report released today by Friends of the Earth groups…' '"The certification of palm oil by the RSPO does not halt deforestation, it does not halt the expansion of damaging oil palm plantations and it does not benefit local communities. Basically, it fails to deal with the causes of the palm oil problems," said Friends of the Earth International Agrofuels Campaign Coordinator Torry Kuswardono from Indonesia. Small but quickly growing quantities of palm oil are being certified by the RSPO. The certification of palm oil is seen by many as a way to make the palm oil industry look "responsible" or "sustainable" ….'

GW None
'… In general, we are not confident about voluntary systems involving businesses. Sometimes they work, and sometimes they do not, and we do not see them as a substitute for binding regulations and accountability. There are times in which there may be an argument for a voluntary system as a first incremental step along the way, but the danger is that a voluntary system negates the potential and opportunity for a mandatory system and that is absolutely what we want to avoid.' 'Particularly in the area of forests and palm oil, we have significant concerns about how the certification schemes are functioning. We do not rule it out. The certification scheme is a very broad term, but there are far too many that do not deliver on the intended aims and vision.' '… Global Witness totally rejects these accusations. We are not funded by multinationals, as a quick look on our website would show, where all our funders are declared (see page 11 of our 2011-12 financial statement) ….' '… A report published by Global Witness today shows how unlikely this is to work. The report - can also assume that the overall success of cross-sector partnerships reinforced their willingness and openness to partner with others.
The path dependency explained above appears to also apply to radical NGOs, especially FOE and GW, but negatively. NGOs that are opposed to the adoption of more collaborative approaches due to the loss of legitimacy and independence of NGOs (Baur & Schmitz, 2012;Burchell & Cook, 2013b) may, as a consequence, reject cross-sector partnerships with MNCs, owing to concerns about the potential for their agendas to be diluted and 'hijacked' by businesses for public relations purposes (Stafford, Polonsky, & Hartman, 1998). For example, the concern for 'selling out' their missions or general avoidance of being associated with a multinational corporation was voiced by FOE and GW.
To sum, those NGOs engaging in cross-sector partnerships were also associated with more positive emotional framing. Both emotional framing and cross-partnership activities intensified over time, demonstrating the role of positive experience and relationships in reinforcing NGO partnership strategies on emotional framing. On the contrary, those who rejected partnerships continued to do so and exhibited increasingly negative emotional framing over time.
These path dependencies present some risks for both radical and reformative NGOs. At the reformative end of the spectrum, NGOs' cooptation might result in the loss of autonomy, reducing their capacity to assume the role as an independent third party next to the state and the market (Ossewaarde et al., 2008) and thus effectively representing the interests of wider society. At the same time, it may result in the loss of public legitimacy of these NGOs and partnerships and even wider private governance initiatives they participate in (Baur & Schmitz, 2012). In other words, there is a risk that reformative NGOs can 'become voices ofrather than watchdogs overofficial agencies, political parties and powerful individuals in global governance' (Scholte, 2004, p. 224). At the radical end of the spectrum, refusal to enter into cross-sector partnerships with the private sector not only means that no direct access to corporate leaders can be gained, thus potentially foregoing the opportunity to bring about change from within and losing a range of significant funding opportunities. It may also lead to the exclusion from significant parts of the recently emerging private governance architecture, including various new-generation multi-stakeholder initiatives. Thus, to a certain extent, there is a risk that some more radical positions become marginalised or even excluded from the organisational field (Laasonen et al., 2012).
The potential co-optation of reformative NGOs and the marginalisation of radical NGOs may limit both the functionality and credibility of multi-stakeholder initiatives. This is significant given that multi-stakeholder initiatives have become essential parts of governance for sustainable development. The credibility and effectiveness of sustainability-related norm-building, rulemaking or standards-setting will suffer if a substantial part of the NGO sector chooses not to engage in these initiatives. From a policymaking perspective, this imbalance creates the need to actively try to promote more radical voices in governance mechanisms to enhance their credibility and balance. This also includes the need for national governments to create conditions under which multi-stakeholder initiatives can be effective governance mechanisms (Gond, Kang, & Moon, 2011).

| CONCLUSION
Our paper makes three distinct contributions to the existing literatures on NGO-business relationships (Baur & Schmitz, 2012;Bendell, Collins, & Roper, 2010;den Hond & de Bakker, 2007;Yaziji & Doh, 2009), cross-sector partnerships (Seitanidi & Crane, 2008;Selsky & Parker, 2005) and framing in the context of NGOs (Luxon, 2019;Pesqueira et al., 2020 Yaziji & Doh, 2009). Second, we demonstrated that NGOs' emotional framing vary depending on their position in the radical-reformative spectrum (Luxon, 2019). Third, as a methodological contribution, we demonstrate how sentiment analysis tools can be usefully applied to the analysis of NGO-business relationships and more specifically to NGO campaigning in the context of MNCs.
We are aware of several limitations affecting our data and methods. The WordStat dictionaries we have used for the sentiment analysis have widespread applicability but could be refined further to capture the specific language and style found in NGO press releases.
Nevertheless, their extensive and successful use in several contexts, many related to ours (Dabic, Colovic, Lamotte, Painter-Morland, & Brozovic, 2016;Lock & Seele, 2016), suggest that this limitation should not be severe. Furthermore, press releases constitute one of a range of engagement mechanisms and contributions to the discourse on corporate responsibility for NGOs, and the method applied in this study, therefore, evaluated one specific dimension (emotional framing)-rather than the entirety-of this discourse. There is also a risk of missing different dynamics that may have been developing in NGOs and sectors other than those that we have included in this study (cf. Idemudia, 2017). Crucially, the trends and patterns identified here may not extend to other types of (e.g., smaller, community-based and service-oriented) NGOs not included in our sample. Therefore, this type of analysis should be replicated with other NGO types. Likewise, the addition of other types of-quantitative and qualitative-data and media and the application of other methods can be expected to generate a more in-depth picture of framing over time and NGO-business relationships more generally.
Future research should include the application of more diverse samples and the fine-tuning of dictionaries used in this study.
Likewise, it could engage more directly with the framing contests between companies and NGOs (MacKay & Munro, 2012), for example, concerning specific norm-building processes such as corporate responsibility standards (Haack, Schoeneborn, & Wickert, 2012). Another promising avenue for future research is a more focused analysis of factors that determine the extent to which international NGOs scrutinise individual sectors and companies. A quantitative analysis of the spotlight effect (Spar, 1998) would contribute towards a better understanding of the dynamics underlying NGO campaigning and shed further light on strengths and limitations of private governance mechanisms in the context of wider governance.
We conclude by highlighting that while we have observed an increasing polarisation in the context of environmental NGOs, it remains to be seen whether and up to which point environmental NGOs will continue to gravitate towards the two extreme positions and how this will impact the private governance architecture.