UvA-DARE (Digital Academic Repository) Credible corporate social responsibility (CSR) communication predicts legitimacy Evidence from an experimental study

Purpose – Companies in challenged industries such as fashion often struggle to communicate credibly with their stakeholders about their social and environmental achievements. Credible corporate social responsibility (CSR) communication, however, has been described theoretically as a predictor of legitimacy for organizations in society, but never proven empirically. The purpose of this paper is to test perceived credibility of a CSR website as a main predictor of input and output (pragmatic, cognitive and moral) legitimacy. Design/methodology/approach – A 2 × 2 between-subjects online experiment with participants recruited from the SoSci Panel ( n ¼ 321) is conducted on an anonymized website of a fashion company. Findings – Credible CSRwebsites result inoutput(cognitiveand pragmatic)legitimacy. However, participation in the CSR decision-making process (input or moral legitimacy) did not matter. Instead, the more subjects accepted the outcome of the CSR communication process, the more they found a company to be legitimate. Research limitations/implications – The CSR communication process on a website is just one specific example. In other settings, such as social media, the role of participation in the CSR communication process will be different. Practical implications – Communicating credibly is a key, particularly in challenged industries, such as fashion. Thus, designing credible communication material matters for legitimacy. Originality/value – The findings for the first time confirm the credibility – legitimacy link in corporate communication empirically. Participation in CSR-related decision-making processes is overrated: the outcome of the CSR communication process is important for stakeholders and their acceptance of a company in society, the participation in the process less. This confirms the idea of CSR as stakeholder expectations management. of meaning making (Basu and Palazzo, 2008). Focusing on one instant of the CSR communication process, i.e., corporate websites, and within the context of the fashion industry, this study attempts to test all three types of legitimacy as consequences of credible CSR communication. To test for cognitive and pragmatic legitimacy, we follow an input-output legitimacy idea where cognitive legitimacy is based on stakeholders ’ expectations and is tested before pragmatic legitimacy (Arnesen, 2017; RQ1 ). We manipulate moral legitimacy, based on its roots in deliberation, through participation in the communication process ( RQ2 ). The study sets out to answer these two research questions:

Websites have predominantly been studied in terms of their potential for dialogue (Kent and Taylor, 1998), consumers ' website attitudes (Chen and Wells, 1999), website performance (Zeithaml et al., 2002) or satisfaction with the website (Bansal et al., 2004). A wide array of research has analyzed CSR on corporate websites; Wanderley et al. (2008) analyzed whether information disclosure on the websites of 127 corporations was influenced by country of origin and/or industry sector. Tang et al. (2015) compared the CSR communications of 50 US and 50 Chinese websites, while Capriotti and Moreno (2007) stretch the affordance of interactivity of corporate websites. In the area of fashion, Da Giau et al. (2016) researched which environmental and social sustainability practices companies in the Italian fashion industry adopt and how they are communicated on their website. However, to date research has not yet investigated websites in terms of their perceived credibility, which is particularly intriguing to study as credible communication can be seen as a predictor of legitimacy (Lock and Seele, 2017).
It is exactly in industries such as fashion that companies face communication challenges and are confronted with legitimacy gaps (Shim and Yang, 2016): stakeholders challenge companies' license to operate because of past scandals or misconduct (Luoma-aho and Vos, 2010). Du and Vieira (2012) explain that controversial industries "may be polemic owing to the goods or services that they provide and/or how they conduct themselves in the process of achieving business objectives" (p. 413). Thus, either industries are considered controversial due to their products or services. Tobacco or alcohol resemble such products as they are by social standards considered sinful, detrimental to health and potentially addictive (Fooks et al., 2011;Du and Vieira, 2012;Cai et al., 2012). Or, industries are seen as controversial when known for participating in socially or environmentally irresponsible business practices or corrupt and unethical behaviors. The oil industry is widely associated with environmental disasters, such as the Exxon Valdez and BP oil spills. Further, it is also perceived to be guilty of negative 4 CCIJ 24,1 social and environmental consequences, such as human rights abuses and neglect of labor and safety standards as well as high CO 2 emissions and the deterioration of water caused by oil production (Woolfson and Beck, 2005).
The (fast) fashion industry resembles a controversial industry because of unsustainable practices within its complex supply chains (Arrigo, 2013). From an environmental perspective, the production processes involve the use of chemical substances posing a threat to the safety of local water and the global transportation of goods resembles an issue in terms of CO 2 emissions (Seuring and Müller, 2008;Caniato et al., 2012;Fletcher, 2013). From a social perspective, it is associated with unethical working conditions and human rights violations (Seuring and Müller, 2008;Arrigo, 2013). In 2013, public pressure increased when the building of a clothing factory collapsed in Bangladesh. Thus, fashion companies are subject to sustainability pressures and attacks by NGOs and the public and are, therefore, especially forced to engage in CSR communication (Pedersen and Andersen, 2015).
Thus, re-and maintaining legitimacy in such a sector is a challenge, particularly because of the high scepticism associated with the CSR communication from these companies (Du and Viera, 2012). But, since this industry is present in everyday life, it sets the proper stage for researching the credibility-legitimacy nexus.

Researching the credibility-legitimacy link
Legitimacy is central to the understanding of organization-public relationships (Heath et al., 2013;Waeraas, 2018), but barely inquired empirically in corporate communication ( Jahn et al., 2017;Bachmann and Ingenhoff, 2016). In neighboring disciplines to corporate communication, most dominantly business management or organizational communication, lots of research have looked at legitimacy and CSR from a qualitative perspective (e.g. Castelló and Lozano, 2011;Golant and Sillince, 2007;Reast et al., 2013;Berry, 2017). However, legitimacy has barely been studied quantitatively as an outcome variable in the context of CSR communication (Crane and Glozer, 2016), and never with regard to credibility.
Past studies have measured legitimacy through the lens of accreditation or firm survival (Ruef and Scott, 1998), in perception surveys (Massey, 2001;Chung et al., 2016), through media coverage (e.g. Deephouse and Carter, 2005), or semantically based on the assumption that positive expressions by stakeholders about a firm are a proxy for legitimacy (Etter et al., 2018). Experimental approaches to measuring legitimacy have been conducted predominantly in the political sciences and social psychology (Arnesen, 2017).
Different types of legitimacy have been differentiated in the literature. Industry legitimacy refers to the idea that entire industries can jeopardize their license to operate in society when their business practice or products have undergone major misconducts, such as the banking industry after the financial crisis or the tobacco sector (Fooks et al., 2013). An issue, on the other hand, is legitimate if the public perceives it as a public concern (Coombs, 1992). Corporate communication researchers have suggested actional legitimacy to look at micro processes and to study specific issue management situations (Boyd, 2009). In the political sciences, decision legitimacy is studied in relation to elections or decisions reached through deliberation (Parkinson, 2003). This is also where the notion of input and output legitimacy originates back from thinkers such as Max Weber (Waeraas, 2018): input legitimacy refers to participation in a decision-making process, output legitimacy looks at the outcome of this process (Mena and Palazzo, 2012;Arnesen, 2017). In a democracy, both need to be in place.
This idea of input and output legitimacy is different from, but also resonates with legitimacy conceptions in management studies. Here, legitimacy of organizations has been defined as the "generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definitions" (Suchman, 1995, p. 574). This idea springs from the notion of the implicit "license to operate" that society grants companies to exist (Donaldson and Preston, 1995). Three classic types of 5 CSR communication organizational legitimacy are distinguished: pragmatic, cognitive and moral. Pragmatic legitimacy refers to the instrumental notion of give-and-take, for instance, a company is a good corporate citizen and therefore customers will buy their products. Pragmatic legitimacy is, thus, connected to an instrumental perspective in CSR theory, where companies act on and communicate their social and environmental responsibilities because they want to achieve a favorable outcome such as higher sales or a better stock market reputation. Strategies to gain pragmatic legitimacy are strategic and exchange based (Nielsen and Thomsen, 2018). Cognitive legitimacy, on the other hand, refers to the normative "taken-for-grantedness" (Suchman, 1995, p. 582) of organizations that is given as long as stakeholders' expectations are satisfied. Thus, cognitive legitimacy is attained if the company's right to exist is commonly accepted by society. Such judgments of cognitive legitimacy are value-and norm-based and are dependent on an overlap between stakeholders' and organizations' norms. However, companies need to know about stakeholders' demands and, thus, ask about their expectations. Thus, "cognitive legitimacy is anchored in the impetus to conform to perception models" (Nielsen and Thomsen, 2018, p. 11). From a political science perspective, both pragmatic and cognitive legitimacy can be regarded as forms of output legitimacy, where cognitive legitimacy needs to be established before pragmatic legitimacy can be achieved (Suchman, 1995;Nielsen and Thomsen, 2018). Hence, stakeholders' norm-based expectations need to be met before they are willing to engage in a (e.g. monetary) exchange.
As a third type, moral legitimacy comes closest to the idea of input legitimacy, as it rests "on judgments about whether the activity is 'the right thing to do' " (Suchman, 1995, p. 579) and is established in a process of moral reasoning (Nielsen and Thomsen, 2018). Here, CSR issues are ideally deliberated with involved stakeholders such that all voices can have a say in the issue and are heard. The notion goes back to the ideas of ethical discourse and deliberative democracy formulated by Habermas, and picked up in political-normative CSR theory (Scherer and Palazzo, 2011).
These three legitimacy types are co-occurring, contingent and formed in a complex process of meaning making (Basu and Palazzo, 2008). Focusing on one instant of the CSR communication process, i.e., corporate websites, and within the context of the fashion industry, this study attempts to test all three types of legitimacy as consequences of credible CSR communication. To test for cognitive and pragmatic legitimacy, we follow an input-output legitimacy idea where cognitive legitimacy is based on stakeholders' expectations and is tested before pragmatic legitimacy (Arnesen, 2017;RQ1). We manipulate moral legitimacy, based on its roots in deliberation, through participation in the communication process (RQ2). The study sets out to answer these two research questions:

RQ1.
To what extent does perceived credibility of CSR websites result in cognitive and pragmatic organizational legitimacy?
RQ2. In how far do participation in the CSR decision and the outcome of the CSR decision-making process influence judgments of organizational legitimacy?

Hypotheses development
While some research found positive effects of corporate credibility on legitimacy ( Jahn et al., 2017;Bachmann and Ingenhoff, 2016), the credibility of the communication tool, here, the website, has not been subject to analysis. To test the credibility-legitimacy relationship and shows that credible CSR communication on a website of a company from a challenged industry leads to cognitive and pragmatic legitimacy, it is hypothesized: H1. Perceived credibility leads to cognitive legitimacy.

6
CCIJ 24,1 Following Suchman (1995, p. 582), cognitive legitimacy "must mesh both with larger belief systems and with the experienced reality of the audience's daily life." Therefore, on the one hand it comes close to the idea of stakeholder expectations management; on the other hand, it is a form of outcome legitimacy, with the stakeholder granting legitimacy if the outcome of the (communication) process is favorable to him/her. Since cognitive legitimacy is operationalized as the perceived agreement with the decision, the input into the decision-making process will logically impact the outcome (Mena and Palazzo, 2012). To test their preferences, respondents could indicate their preference for a CSR topic that was to be included in the CSR website (topic choice). Therefore, stakeholders that stated their preference for a topic which was finally also chosen by the company will likely agree more with the decision (decision agreement). Pragmatic legitimacy entails "an exchange of benefits between companies and stakeholders to create a symbiotic relationship" (Berry, 2017, p. 126). This describes a give-and-take situation where the stakeholder receives something in return for some investment. Thus, we view pragmatic legitimacy as given if a stakeholder is willing to buy a product from the company, as this counts as a direct exchange and is one of theif not themost researched outcome variable in an instrumental, promotion-based CSR perspective (Nielsen and Thomsen, 2018): H3. Topic choice has a positive influence on decision agreement. The more the topic choice is in line with the outcome, the higher the agreement to the decision.
Since "pragmatic, moral, and cognitive legitimacy co-exist in most real-world settings" (Suchman, 1995, p. 584), and cognitive legitimacy is a basis for pragmatic exchanges between stakeholders and companies, we assume that the preference for a topic will also impact pragmatic legitimacy positively: H4. Topic choice has a positive influence on purchase intention. The more the topic choice is in line with the outcome, the higher the intention to purchase.
Furthermore, the more a stakeholder's expectations are met as tested here in terms of his/her agreement with the decision, the higher we also assume their intention to purchase a product of the company will be, thus testing cognitive legitimacy as an antecedent to pragmatic legitimacy: H5. Cognitive legitimacy has a positive influence on pragmatic legitimacy. Arnesen (2017, p. 147) found that "[t]he respondents clearly react more negatively to the decision when the outcome differs from their stated preference." Applied to our context, we can test whether stakeholders' stated topic preferences was more important for their legitimacy perceptions than receiving credible CSR communication. This is also in line with the idea that CSR communication meets a "cynical public" (Illia et al., 2013) that seeks personal gains and overall skepticism toward this communication form (Waddock and Googins, 2011): H6. The effect of topic choice on legitimacy is higher than perceived credibility on legitimacy.
The idea behind moral legitimacy as further developed by political CSR scholars (Scherer et al., 2013) is that it emerges from a communication process that is credible and oriented toward consensus; thus, participation in the decision-making process (input legitimacy) as propagated in deliberative democracy studies is key for this type of legitimacy. Since, it is difficult to measure this latent construct, this study approximates a situation of moral legitimacy. We manipulate participation in a CSR communication process as on the one hand participating in the decision and on the other hand stating a preference in the topic of the decision. This follows approaches to studying the perceived decision-making influence of citizens on the favourability of political decisions deliberative democratic settings in the political sciences (Arnesen, 2017). The basic idea of Arnesen's (2017) study is 7 CSR communication to test what is more important to subjects when they can decide to spend money on an issue: that they can participate in the outcome or that the outcome is in line with their preference. Given that political CSR theory is rooted in deliberative democracy theory, this seems an apt approach to approximate the normative concept of moral legitimacy, because it simulates that stakeholders can voice their preferences and, thus, deliberate about CSR issues (Nielsen and Thomsen, 2018).
We approximate conditions of moral legitimacy by manipulating stakeholders' participation in the decision-making process of the CSR communication situation. Political CSR theory assumes, based on Habermasian deliberative democracy, that participation in a communication process is a normative foundation of all deliberation (Scherer and Palazzo, 2011). A deliberativeand credible -CSR communication process, thus, can lead to moral legitimacy (Berry, 2017). Therefore, the following hypotheses are proposed: H7. Participation in the decision-making process influences cognitive legitimacy positively.
H8. Participation in the decision-making process influences pragmatic legitimacy positively.
Additionally, we assume that the perceived credibility of the website can be affected by existing attitudes about the industry, CSR in general and participants' prior knowledge about the product. Due to prominent cases of corporate hypocrisy, many consumers dismiss CSR measures as window dressing or greenwashing (Connors et al., 2017). Therefore, in general skepticism toward an organizations' CSR claims exists (Brown and Krishna, 2004), which is why the attitude toward CSR can influence reception (Obermiller and Spangenberg, 1998). At the same time, the fashion industry is a challenged industry that struggles with negative reputation. Since CSR and its communication are industry dependent (Fifka, 2013), we controlled for participants' attitudes toward the fashion industry. Last, an individuals' involvement in the company's product, in our case clothing, is further expected to affect credibility judgements, because consumers evaluate information about a high involvement product in line with their existing attitude (Bloch and Richins, 1983). Thus, we included the control variables CSR attitude, industry attitude and clothing interest in our model. The final conceptual model is depicted in Figure 1.

Experimental design and participants
The study employs a 2 between-subjects (participation: participating vs not participating) × 2 between-subjects (topic choice: preferred choice vs not preferred choice) experimental design to test the hypotheses. German participants were recruited via the SoSci Survey Panel (Leiner, 2014) between the December 10, 2017 and January 10, 2018. A total of 423 participants clicked on the link sent to them, but only 328 filled out the questionnaire completely. An additional seven participants were deleted, since they failed to answer an attention check question correctly. This resulted in 321 participants (56 percent female, 44 percent male; age M ¼ 40.63, SD ¼ 16.25). Overall, the sample had a slightly higher level of education than the German population (2 percent secondary school diploma, 11 percent junior high school diploma, 31 percent high school diploma, 19 percent Bachelor's degree, 29 percent Master's degree, 6 percent PhD, 1 percent not completed and 2 percent not specified). Table I shows the distribution of participants to experimental groups.

Stimuli
The stimulus material of the study consists of a CSR website of an anonymized clothing label. On the website, the label presents three of their current CSR measures; one about the increase of organic cotton, the second about the promise to increase transparency about the company's supply chain and the third on fight of child labor. The design and content of the website was adapted and translated from the English CSR website of the clothing label C&A.

Participation
Participation serves as a between-subjects variable and takes two conditions, participating vs not participating. Participation was manipulated by giving participants the choice between two CSR measures, one topic about support for women and another about recycling of textiles. Subjects were asked to indicate which CSR topic they preferred to be added to the CSR website of the label. In the participation condition, the participants received the question: "You are a regular customer of the clothing label. The label wants to invest an additional 100,000 Euro in a new CSR measure and would like to know your opinion. As a regular customer, you have a say in in this decision. Which topic do you prefer to be included in the website?". Thus, in the participating condition, participants could actively take part in the decision-making process since their preference for a new CSR measure was taken into account by the label in its decision. In the not participating condition, the question wording was changed to: "You are a regular customer of the clothing label. The label wants to invest an additional 100,000 Euro in a new CSR measure and would like to know your opinion. Which topic do you prefer to be included in the website?" Here, participants stated their preferred CSR measure, but were not told that their choice would impact the decision of the label and thus could not actively take part in the decision-making process, but only indicate their preference.
Group 4 (59) Notes: The unequal number of participants per group is due to the study design: overall, in both the participating and not participating condition, participants selected the measure "textile recycling" (62 percent) almost twice as much as "support for women" (38 percent). They were then randomly presented with either "textile recycling" or "support for women" as the chosen measure by the clothing label, which resulted in unequal groups

9
CSR communication which CSR measure to include on the website is align with the participant's preferred choice in the participation manipulation. Thus, subject's preferring the support for women (textile recycling) measure, will receive the message "The label has decided to include the measure support for women (textile recycling) on their website" and are shown how the new CSR-measure is presented on the website. In the not preferred choice condition, the final decision of the clothing label deviates from the participant's preferred choice. This means participants who chose to include support for women (textile recycling) receive the message "The label has decided to include the measure textile recycling (support for women) on their website," again showing how the CSR-measure is presented on the website.

Measures
Perceived credibility assesses how credible participants rated the clothing label's website by judging the content in the four dimensions understandability, truth, sincerity and appropriateness. It was measured by the validated 16-item PERCRED scale by Lock and Seele (2017) (see Appendix 1) on a five-point Likert-scale (1 ¼ "Strongly disagree"-5 ¼ "Strongly agree"). The reliability of the scale was excellent and the averaged to form a composite score for analysis (M ¼ 3.36, SD ¼ 0.64, α ¼ 0.90).
CSR attitude serves as a control variable to estimate participant's existing attitudes of CSR practices. CSR attitude is measured by three items of the adapted consumer-skepticism scale by Obermiller and Spangenberg (1998) on a five-point Likert-scale (1 ¼ "Strongly disagree"-5 ¼ "Strongly agree") (see Appendix 1). The reliability of the scale was good and items were averaged to form a composite score for analysis (M ¼ 3.51, SD ¼ 0.87, α ¼ 0.81).
Industry attitude was assessed to control for participant's existing opinions about the clothing industry. It was measured by five adjectives on a five-point semantic scale. The reliability of the scale was good and averaged to form a composite score (M ¼ 3.62, SD ¼ 0.65, α ¼ 0.81).
Clothing interest controls for a person's interest in fashion matters and was measured by the adapted four-item "Interest and Involvement" scale by Mehta and Sivadas (1995) on a five-point Likert-scale (1 ¼ "Strongly disagree"-5 ¼ "Strongly agree"). The reliability of the scale was good and averaged to form a composite score (M ¼ 2.10, SD ¼ 0.87, α ¼ 0.88).
Decision agreement is a dependent variable and assesses to what extent the participants agree with the label's decision regarding its final decision on which of the two proposed CSR measures to include on the website. It is measured by the item "To what degree do you find this decision acceptable?" on a five-point Likert scale (1 ¼ "Strongly disagree"-5 ¼ "Strongly agree") (M ¼ 3.62, SD ¼ 1.14).
Purchase intention is the second dependent variable and asks the participant to indicate the likelihood to purchase products from the label. It is assessed by the three-item, five-point semantic scale by Yi (1990) (Appendix 1). The reliability of the scale was excellent and averaged to form a composite score (M ¼ 3.59, SD ¼ 0.93, α ¼ 0.90).

Procedure
Respondents accessed the experiment by clicking on a link sent to them by the SoSci panel on December 10, 2018. On the opening page, respondents were informed that the study is conducted by the Amsterdam School of Communication Research and investigates their perception of a CSR website of an anonymous fashion label. Furthermore, they were assured of their anonymity, provided with contact details and asked to give their informed consent before starting the questionnaire.
First, respondents answered the control variables industry attitude and interest in clothing. They were then shown the CSR website of the label. After reading the website, respondents were asked to evaluate the website by answering the items of the perceived credibility scale. They were then randomly assigned to the two participation conditions in 10 CCIJ 24,1 which they indicated their preferred CSR-measure. This was followed by a distraction exercise before continuing the questionnaire for which participants were able to provide their e-mail address for the chance of winning one of five Amazon gift vouchers. Based on their topic choice of CSR-measure, participants were then randomly assigned to either the preferred condition or the not preferred condition. Lastly, they answered the items of the dependent variables decision agreement and purchase intention as well as the items of the CSR attitude scale.
After finishing the questionnaire, respondents were thanked and debriefed that the true intention of the questionnaire was to measure the influence of participation and topic choice on decision agreement and intention to purchase. Further, they were told that the website was inspired by the CSR website of the label C&A, but was not involved in the study.

Randomization check
Randomization checks yielded no significant differences across the groups regarding the demographic variables age and gender and regarding the control variables interest in clothing, CSR attitude and industry attitude (see Table AI). This means that participants were equally divided across groups, and it can therefore be expected that these variables are not responsible for significant differences between groups.
Model specification and fit. To test the hypothesized theoretical model, we specified a path model in AMOS in which we included all constructs as manifest variables. The complete path model is depicted in Figure 2. To evaluate model fit, the following criteria were used: Model χ 2 -test, in which a non-significant p-value represents acceptable model fit (Kline, 2011), CFI, which should range between 0.90 and 1.00 for acceptable model fit (Bentler, 1990), and RMSEA, with a cut-off value ⩽ 0.05 (Mulaik, 2009). The model converged well and fit of the model was excellent ( χ 2 (8) ¼ 9.029, p ¼ 0.340, CFI ¼ 1.00, RMSEA ¼ 0.02, CI {0.00; 0.07}).
Overall, the theoretical model could explain about 14 percent (R 2 ¼ 0.14) of the variance of cognitive legitimacy, measured by decision agreement, and about 39 percent (R 2 ¼ 0.39) of the variance of pragmatic legitimacy, measured by purchase intention.
Hypotheses testing H1 states that perceived credibility leads to cognitive legitimacy. Cognitive legitimacy was assessed in the experiment by decision agreement. The results show that there was a moderate positive effect from perceived credibility on decision agreement ( β ¼ 0.24, p o0.001). Thus, H1 can be accepted. H2 states that perceived credibility leads to pragmatic legitimacy, measured by the intention to purchase. Perceived credibility exerts a moderate

11
CSR communication effect on purchase intention ( β ¼ 0.48, p o0.001); thus, we accept H2. Overall, perceived credibility of the CSR website leads to both cognitive and pragmatic legitimacy.
H3 concerns the effect of topic choice on decision agreement and states that the more the topic choice is in line with the outcome, the higher the agreement to the decision. In the path model, the not preferred choice was given the value "1," while the preferred choice was given the value "2." Thus, the moderately positive effect of topic choice on decision agreement ( β ¼ 0.29, p o0.001) supports H3. Therefore, we conclude that participants agreed more with the decision of the fashion company when it matched their own preference.
Regarding H4, which stated that topic choice (stated preference) has a positive influence on purchase intention, no significant effects were found and H4 is therefore not supported. This means that topic choice did not have a direct effect on purchase intention, and does therefore not predict pragmatic legitimacy. However, the indirect effect of topic choice on pragmatic legitimacy through cognitive legitimacy was significant ( β ¼ 0.16, p o0.001). Therefore, the effect of preferred topic on purchase intention was fully mediated by decision agreement, further emphasizing that cognitive legitimacy leads to pragmatic legitimacy. Thus, the hypothesis can only partially be confirmed.
We hypothesized that cognitive legitimacy has a positive effect on pragmatic legitimacy (H5). The results show a moderately positive effect from decision agreement on purchase intention ( β ¼ 0.31, po0.001), confirming H5. Thus, agreement with the decision of the fashion company leads to higher intention to purchase its products, thus pragmatic legitimacy.
H6 states that the effect of topic choice on legitimacy is higher than the effect of perceived credibility on legitimacy. Comparing the effect sizes, the hypothesis can be confirmed for cognitive legitimacy, since the effect size of topic choice on decision agreement is higher ( β ¼ 0.29, p o0.001) than the effect of perceived credibility ( β ¼ 0.24, p o0.001).
Regarding the effect of preferred topic on pragmatic legitimacy, the effect of topic choice on purchase intention is not significant.
Lastly, for H7 and H8, we expect a positive effect of participation in the decision-making process on both cognitive and pragmatic legitimacy. For both, the results show a non-significant effect of participation. Hence, participation in the decision-making process on attributing legitimacy did not significantly differ, which leads us to reject H7 and H8 (Table II).

Additional results
The control variables CSR attitude, industry attitude and interest in clothing were introduced in the model to control for their impact on perceived credibility (Lock and Seele, 2017). As discussed previously, they did not significantly differ across experimental groups and therefore do not impact the interpretation of the results. Interestingly, however, all control variables exert a significant effect on perceived credibility and account for 18 percent (R 2 ¼ 0.18) of its explained variance. CSR attitude exerted a Hypothesis Confirmed  moderately negative effect on perceived credibility, meaning that participants that were more skeptic towards CSR practices of organizations also perceived the website as less credible ( β ¼ −0.24, p o 0.001). The same trend was found with respect to industry attitude. Those who had a worse opinion of the clothing industry also perceived the website of the label as less credible ( β ¼ −0.23, p o 0.001). Interest in clothing, on the other hand, showed a small positive effect on perceived credibility ( β ¼ 0.15, p ¼ 0.003), meaning that those more interested in fashion matters evaluated the website as more credible.
Discussion and contribution: confirming the credibility-legitimacy link In general, our results suggest that the credibility of corporate communication tools cannot be overrated. Although there is no one best way to communicate CSR as invoked by some authors in corporate communication, we can suggest with confidence that communicating credibly is key. Attending to source credibility alone ( Jackob, 2008) does not do justice to complex organization-stakeholder relationships managed through CSR communication (Heath et al., 2013), but the credibility of the tools needs to be in the focus. The more credible the CSR website of a firm, the more likely it is that pragmatic and cognitive legitimacy perceptions are evoked in stakeholders. Thus, cognitive and then pragmatic legitimacy results from a credible, output-based communication process. This main finding confirms the theoretical link established by the normative CSR theory (Scherer and Palazzo, 2011;Seele and Lock, 2015). For the political CSR theory, this study shows that credible communication is key "for the continuous flow of resources and for securing the sustained support of the organization's constituencies" (Scherer et al., 2013, p. 262), for which legitimacy is a necessary precondition. But also instrumental conceptions of CSR can benefit from our insights: we show that pragmatic outcomes of CSR communication processes such as purchase intentions are enhanced with credible CSR communication. Contributing to a more constitutive CSR perspective, we confirm that communication is indeed key to legitimation processes and that perceptions of CSR communication are dependent on the context. Credibility perceptions are moderately negatively influenced by participants' prior attitudes toward CSR and toward the industry of the company, while the more interested subjects were in fashion, the more credible they found the website to be.
The type of legitimacy matters include: output, i.e. cognitive and pragmatic, and legitimacy was more important to stakeholders than input, i.e. moral, legitimacy. The more participants' preferences were met, the more they found a company to be legitimate. This cognitive legitimacy is a precondition for pragmatic legitimacy perceptions. Thus, engaging stakeholders succeeds if their preferences are taken into account (Lim and Greenwood, 2017). This confirms the idea of CSR as stakeholder expectations management (Podnar and Golob, 2007), particularly in challenged industries such as pharmaceuticals (O'Riordan and Fairbrass, 2008) or fashion, where companies try to respond to stakeholders' expectations because of the public spotlight and past scandals. Regarding cognitive legitimacy, meeting stakeholders' preferences was even more important than credible CSR communication. However, this was not the case for pragmatic legitimacy. Thus, isomorphic adaptation to changing stakeholder expectations and attitudes can be a key for re-and maintaining legitimacy (Scherer et al., 2013).
Against political-normative approaches to CSR communication, our results show that participation in the CSR communication and decision-making process does not predict legitimacy. Whether the respondents had an influence on the decision which CSR topic should be treated on the website, did not significantly impact how they perceived the decision. As Arnesen (2017) puts it, "output legitimacy trumps input legitimacy." The idea of deliberation as a key to reaching an understanding between different parties in a 13 CSR communication discourse being the backbone of the deliberative democracy theory (Habermas, 1996) can, however, still hold true for CSR communication, but such understanding does not necessarily lead to more (or less) perceived legitimacy. Thus, facilitating participation in a CSR communication process should not be a goal in and of itself, because there is no direct link to legitimacy. Rather should active stakeholder participation be encouraged because such participation results in better decisions (Arnesen, 2017). Thus, moral legitimacy can be regarded more as a procedural legitimacy component that defines legitimate communication processes for reaching better legitimacy outcomes rather than an end in itself. Therefore, the emphasis of participation in political CSR communication might be exaggerated in practice. The source of legitimacy, as tested here, is clearly the credibility of the corporate communication tool, not participation. This finding, however, does not entail that stakeholder engagement is to be neglected; rather, credible communication is its basis.

Conclusions
This study adds to the debate on public relations in post-truth times by showing that stakeholders regard credible CSR communication as more legitimate than non-credible information. It provides empirical evidence that perceived credibility of a CSR website is a main predictor of corporate legitimacy and, thus, fills an empirical gap. We also extend current corporate communication and CSR communication theory by finding that participation in CSR-related decision-making processes is overrated: it is the outcome of the CSR communication process that is important for stakeholders and for their acceptance of a company in society, not so much the participation in the process. Hence, cognitive and pragmatic legitimacy are key, while moral legitimacy is regarded as making decisions better, but not as making them more acceptable. Of course, the CSR communication process of this experiment is just one specific example of constructing a CSR reality. In other settings, such as social media and including other stakeholders such as ordinary citizens, the role of participation in a CSR communication process will be different (Etter et al., 2018;Romenti et al., 2014). But when it comes to participation of stakeholders in a two-way communication setting involving popular channels such as a website, credibility weighs more for perceived legitimacy than "having a say" in an issue.
These findings hold important practical implications: to successfully manage organization-stakeholder relations, designing credible communication material matters for legitimacy . Referring to understandability of the contents, truthful information, sincere communication and an appropriate context for different stakeholders can be key points for consideration. Such credible communication is not only desirable from an ethical perspective, but may also positively impact on organizational performance and the flow of resources. When managing credibility perceptions of a company's CSR communication, corporate communication practitioners are advised to attend to the company in its broader context: its industry and current issues, and also keep a close eye on stakeholders' preconceptions. In this vein, flexibly adapting to stakeholder expectations is a key for companies to maintain legitimate in the eyes of their constituents.
Limitations and directions for future research These findings do not come without limitations. Our results are based on a German sample representing potential customers of an anonymized label without prior reputation. Future studies should, therefore, test our model on international and different stakeholders such as business partners and employees and with existing fashion labels to increase validity. Furthermore, including the flip side of credibility perceptionsgreenwashingas a concept in further experiments would be worthwhile 14 CCIJ 24,1 in terms of discriminant validity. Here, it would be relevant to take attitudes toward the business, not only the industry, into account. Due to the controversial nature of the fashion industry, research should also investigate the role of credible CSR communication in other industries not immediately associated with questionable practices. Further, our study assessed the perceived credibility of a CSR website which did not allow for further interaction between the users and the organization. As a next step, research could be conducted on a different communication channel, such as on social media that allows for interactivity. This is especially relevant as it would put the communicative process in focus, which is relevant in a deliberative democracy to reach consensus and give legitimacy to actions.