Maximising Business Returns to Corporate Social Responsibility Communication: An Empirical Test

Based on a classic conceptual model of corporate social responsibility (CSR) communication, we developed an empirical research study to test how several aspects of CSR message content (i.e., issue importance, impact, motives, fit, commitment) are associated with external support responses (i.e., purchase, advocacy). We also tested the moderating role of stakeholder&#8208; and company&#8208;specific factors (i.e., issue support and industry, respectively) in the proposed model. Data were collected from 302 participants who evaluated the same CSR information displayed in the websites of a fictitious bank and a fictitious restaurant chain. The findings suggest that better perceptions on how the CSR message reinforces issue importance, corporate CSR impact and altruistic motives lead to higher purchase and advocacy intentions. CSR fit is related only to advocacy, while CSR commitment does not have any significant impact on participants&#8217; responses. Some new interdependence relationships are also identified among issue importance, motives, fit, and commitment. The moderating role of issue support and industry is confirmed.

The research goal of this paper is to fill these two gaps in the literature by testing the conceptual model of Du et al. (2010) in an empirical setting. More precisely, the model helps us identify which aspects of the message (i.e., issue importance, CSR impact, CSR motives, CSR fit, CSR commitment) are more effective in generating positive consumer behaviour (i.e., purchase, advocacy). Also, as suggested by Du et al. (2010), we explore the moderating role of two contingency factors related to the consumer (i.e., issue support) and the company (i.e., industry) in the model. Our main hypothesis is that manipulating all these variables may not be equally effective. Thus, companies would benefit from knowing which aspects of their CSR messages should be especially attended to achieve better communication results.
In doing so, the paper contributes to previous literature by complementing our knowledge on the efficacy of CSR communication using Du's et al. (2010) model. While previous papers have failed to provide integrative models that allow researchers and practitioners to understand consumer behavioural responses completely, in this paper, the authors demonstrate that consumer responses to CSR communication are affected by multiple variables and that interaction effects also exist among them. By considering multiple variables simultaneously, researchers can design conceptual and empirical models with greater predictive and explanatory power of the effectiveness of CSR communication. In this regard, the paper identifies new relationships among variables related to the CSR message content that previous researchers have neglected to study. As it will be explained in detail in this paper, some examples include the relationships that are manifested in the research between CSR fit and issue importance, CSR fit and CSR commitment, or CSR motives and CSR commitment.

| What to communicate: CSR message content
The persuasiveness of a communication can be increased easily and dramatically by paying attention to the message (Darley & Smith, 1993). Previous research has demonstrated that the manipulation of key informational content within the message influences consumer responses because it affects the level of scepticism to both the message's believability and attributions for the company's motivations for involvement in the cause, among other key issues (Forehand & Grier, 2003). Du et al. (2010) identify as many as five factors that a company can emphasise in its CSR message to improve stakeholders' perceptions, attitudes, and behaviours ( Figure 1). First, issue importance refers to the prominence that the company gives to CSR and the social cause in its communicational messages (Russell & Russell, 2010).
Second, CSR impact is defined as the output side of the company's CSR endeavour, that is, its societal impact, or the actual benefits that the company have accrued (or will accrue) to the target audience of a social cause (Du et al., 2010). Third, CSR motives refer to the reasons and intentions that are attributed to the company when it engages in CSR, which could be altruistic/intrinsic motives or egoistic/extrinsic motives (Marín & Ruiz, 2007). Fourth, CSR fit is defined as the overall relatedness of the company and the social cause it supports (Bigné, Chumpitaz, & Currás, 2010;Pracejus, Olsen, & Brown, 2004).
Fifth, CSR commitment reflects the amount of input provided by the company to the social cause, the durability of the association and the consistency of input (Dwyer, Schurr, & Oh, 1987).
In terms of consumer behaviour, Du et al. (2010) report that these five variables can affect purchase intentions and behaviour, or the consumer commitment to continue purchasing the same goods or using the same service from a company over time (Oliver, 1997).
These variables also are associated with consumer advocacy behaviour (Du et al., 2010). In this regard, advocacy refers to positive word-of-mouth, which represents informal communication directed at other consumers regarding products, services, and companies (Westbrook, 1987). Its importance lies in the effect these recommendations have on people close to the consumer because they may encourage them to also show interest in the company and buy its products or services (Oliver, 1997). Although Du et al. (2010)  loyalty as a third external outcome to be evaluated when consumer responses are explored, we did not include this variable in the model. This decision was based on an extensive amount of previous literature that defined loyalty as a reflection of repeated purchase and advocacy behaviours instead of defining it as an independent construct (Dick & Basu, 1994;Homburg & Giering, 2001;Oliver, 1997).
Including loyalty in empirical models along with purchase and advocacy could be problematic in terms of content, convergent, and discriminant validity.
For instance, a CSR message that highlights the importance of a CSR cause (i.e., issue importance) provides consumers with social topic information (Pomering & Johnson, 2009), thereby increasing their awareness of the cause and enabling them to effectively process CSR appeals (Auger, Burke, Devinney, & Louviere, 2003).
Otherwise, consumers may lack sufficient prior knowledge of the cause, which would make it less accessible and harder to recall when evaluating the message and the company (Tybout, Sternthal, Malaviya, Bakamitsos, & Park, 2005). Adequately informing consumers about a social problem will enable them to draw on those associations to activate socially evaluative criteria, allowing the ease with which such information comes to mind to serve as the basis for judgement and improvement of responses to CSR messages (Tybout et al., 2005).
Another relevant aspect relates to the CSR impact that a company claims to accrue to the target audience of a social cause (Du et al., 2010). Several researchers believe that focusing the message on the output side and the results of the collaboration in the cause derives in better consumer responses (Du et al., 2010;Gregory-Smith, Manika, & Demirel, 2017;Wood, 1991). As suggested by Pracejus et al. (2004), consumers are especially interested in being able to evaluate the true level of a company's CSR involvement, which is confirmed by the finding that CSR donation amount influences persuasion effects and has a significant impact on consumer choices (Pomering & Johnson, 2009).
Consumers reportedly are sceptical of a company's CSR claims due to attributions of self-interest to the company's activities (Forehand & Grier, 2003). Principles followed by companies to motivate their CSR involvement are coded in three categories (Maignan & Ralston, 2002): value driven (i.e., CSR is presented as being part of the company's culture or as an expression of its core values), stakeholder driven (i.e., CSR is presented as a response to the pressure and scrutiny of one or more stakeholder groups), or performance driven (i.e., CSR is introduced as a part of the company's economic mission as an instrument to improve its financial performance and competitive posture).
The literature agrees that consumers prefer companies that show altruistic/intrinsic (i.e., value-driven) motivations to support a cause over comparable companies that form alliances with causes only to generate sales or elude conflicts (i.e., performance-driven or stakeholder-driven companies to which are attributed extrinsic motivations) (Barone, Miyazaki, & Taylor, 2000;Nan & Heo, 2007).
More precisely, the perception of a company's intrinsic motives suggests recognition of a certain amount of transparency, which increases perceived sincerity . In contrast, behaviour attributed to extrinsic motives is perceived as dishonest and misleading. It suggests that the cause would not have been supported without a reward and therefore appears opportunistic. In the context of CSR communication, extrinsic attributions should induce a perception of self-serving motives and therefore weak sincerity and consumer behavioural responses .
Along this line, a fourth relevant aspect of CSR message content relates to CSR fit (Du et al., 2010). CSR fit has a dual nature (Lafferty, Goldsmith, & Hult, 2004;Trimble & Rifon, 2006) because consumers can perceive either image or functional fit when analysing the collaboration between the company and the cause.
Although image fit refers to the holistic, symbolic, and peripheral judgement of company identity and its relatedness to the cause, functional fit pertains to the compatibility of the type of product/ service marketed by the company, and the type of social cause supported (Bigné et al., 2010). Roughly speaking, the literature recognises that consumers evaluate high-fit collaborations more positively than low-fit activities (Aqueveque, Rodrigo, & Duran, 2018;Skard & Thorbjornsen, 2014;Weeks, Cornwell, & Drennan, 2008).
As explained by Dawkins (2004), for credibility, the causes companies support must be seen to fit with their business, and their corporate behaviour as a whole must be seen to be consistent.
Otherwise, corporate CSR messages risk being regarded as a smokescreen for unethical behaviour. Benoit-Moreau and  confirm that the perceived congruence between the company and the cause reinforces the impact of CSR communication on brand equity. Because brand equity includes aspects related to purchase, advocacy, and loyalty towards the company (Lai, Chiu, Yang, & Pai, 2010), CSR fit is demonstrated to have a direct impact on consumer external outcomes.
Finally, a long-term commitment to CSR across different operating activities of the company, as opposed to more short-term and opportunistic promotional CSR, also may provoke less scepticism and improve purchase and advocacy responses (Pomering & Johnson, 2009). We expect that information which establishes a company's long-term commitment to CSR will be diagnostic in CSR messages, therefore improving responses. For this purpose, CSR commitment can be demonstrated by referring to the amount of support given to the cause, the durability of the support and its consistency over time (Dwyer et al., 1987).
Based on these ideas, we propose two research hypotheses to test the effects of CSR message content on purchase and advocacy outcomes. They are: Electronic copy available at: https://ssrn.com/abstract=3587840 Although Du et al. (2010)  First, if CSR motives are altruistic, they do not derive from a short-term goal such as the search for direct economic benefits or pressure from stakeholder groups. It then is expected that altruistic motives will lead consumers to perceive the company's commitment to the cause as more lasting because the motivation for collaboration comes directly from the company's value system (Maignan & Ralston, 2002) that, as proposed by strategic management researchers, is stable over time (van Rekom, van Riel, & Wierenga, 2006 CSR fit also can be expected to have significant positive impacts on the attribution of CSR motives, consumer perceptions of corporate CSR commitment, and perceptions of the issue importance reported in CSR messages (Bigné et al., 2010;Menon & Kahn, 2003;Simmons & Becker-Olsen, 2006).

For instance, CSR fit is important because it affects consumers'
CSR attributions (Menon & Kahn, 2003;Simmons & Becker-Olsen, 2006). Consumers first will attribute CSR activities to intrinsic motives and then correct this inference if they allocate sufficient processing capabilities and engage in more effortful elaboration by considering alternative, contextual factors (Du et al., 2010). Low CSR fit, owing to the lack of logical connection between a cause and a company's core business, is likely to increase cognitive elaboration and make extrinsic motives more salient, thereby reducing consumers' positive responses to a company's CSR message. Therefore, when a company does not have a good natural fit with the cause it supports, it should elaborate on the rationale for its collaboration to increase perceived fit (Bigné et al., 2010).
Additionally, when congruence is high, it is expected that the degree of CSR commitment of the company is greater because consumers will consider that for the company it is easier to collaborate with that cause than with a CSR activity that has nothing to do with the company's core business; this fact would increase the probability that the company commits to the cause in the long term. When congruence is high, it also is expected that the consumer will perceive more easily and strongly the importance that the company gives to the cause in the message. Specifically, if the company supports a cause very close to its core business, consumers will consider that it is a cause that is very important for the raison d'être of the company.
Based on these ideas, we propose the following research hypothesis:

| Moderators of communication effectiveness
Effectively communicating CSR is not a straightforward task. In addition to message variables, which are directly controlled by the company, the effectiveness of CSR communication also is likely to be influenced by extraneous variables that are out of the company's control, such as stakeholder-specific factors (e.g., support to the company's CSR domain) or company-specific factors (e.g., news eventually generated in the industry) (Du et al., 2010;Pomering et al., 2013; Sen Bhattacharya, 2001). For the purpose of corroborating this idea, we take two stakeholder-and company-specific variables as examples to test the existence of moderating effects in our empirical model.
On the one hand, we explore the role of issue support. Issue support refers to the tendency for consumers to purchase products/services that they perceive to have a positive (or less negative) impact on the society or to use their purchasing power to express current social concerns Roberts, 1995). Thus, issue support refers to the relevance or personal importance that CSR has to a consumer, based on the person's needs, values, and inherent interests (Zaichkowsky, 1985). Podnar and Golob (2007)  Research has shown that information perceived as self-relevant (vs. non-relevant) elicits voluntary attention (Petty, Unnava, & Strathman, 1991). Because issue support reflects personal needs and values, all else being equal, CSR information on initiatives that consumers deem important or personally relevant is more likely to break the media clutter and be more effective (Du et al., 2010). Thus, we expect that for people showing high issue support, our empirical model will work better and CSR message content will have stronger effects on support behaviours (i.e., purchase and advocacy). Therefore, a fifth research hypothesis is proposed:

H5: Issue support moderates the relationship between perceptions of the CSR message content and external communication outcomes.
On the other hand, we propose that industry also may moderate the support behaviours that derive from perceptions of CSR message content (Du et al., 2010). Under the light of the institutional theory (Deegan, 2002), previous research has demonstrated that significant differences exist in the effectiveness of CSR communication between high-and low-profile industries (Hackston & Milne, 1996;Patten, 1991) because diverse sectors face different challenges to communicate CSR and encourage stakeholders' support (Aqueveque et al., 2018;Esrock & Leichty, 1998;Maignan & Ralston, 2002;Peattie, Peattie, & Ponting, 2009).
As opposed to low-profile industries, high-profile industries face greater stakeholder pressures, are exposed to higher visibility, and receive greater scrutiny from stakeholders, who are especially critical of the CSR communication coming from companies (Roberts, 1992). For instance, first-time environmental award announcements generally are associated with greater increases in the market value of companies, although smaller increases are observed for companies in environmentally dirty industries (i.e., high-profile industries), possibly indicative of market scepticism (Klassen & McLaughlin, 1996).
Although classifications are to an extent subjective and ad hoc, most scholars identify companies in basic industries as high profile.
In the context of our research, the financial and banking industry also is considered a high-profile industry because it recently has attracted great attention due to the latest economic recession that has especially threatened these companies worldwide (Pérez, García de los Salmones, & López, 2015). This circumstance generates new forms of coercive pressures in exchange for continued legitimacy and can make CSR communication less effective.
Also, studies demonstrate that the effectiveness of marketing strategies for a particular service depends on service type (i.e., hedonic, utilitarian) (Andreu et al., 2015;Hill, Blodgett, Baer, & Wakefield, 2004;Stafford & Day, 1995). Hedonic consumption reflects multi-sensory, fantasy, and emotive aspects of consumer experience, whereas utilitarian consumption focuses on functional consequences (Jiang & Lu Wang, 2006). Hedonic services provide consumers with values such as excitement and playfulness (e.g., restaurants). Utilitarian services, in contrast, provide consumers with functional utilities or solve practical problems (e.g., banking).
Thus, researchers argue that consumers evaluate utilitarian products primarily using cognitive criteria, whereas they evaluate hedonic products on affective issues (Kempf, 1999).
Based on these ideas, we propose that participants in our research will evaluate CSR communication coming from a bank differently from CSR communication associated with a restaurant chain. More precisely, people evaluating banks and restaurants are expected to give different importance to diverse aspects of CSR message content such as issue importance, CSR impact, CSR fit, CSR motives and CSR commitment, and therefore their responses in terms of purchase and advocacy behaviours will differ. Thus, the last research hypothesis is: H6: Industry moderates the relationship between perceptions of the CSR message content and external communication outcomes.

| Research design and sample
We conducted a quantitative study based on interviewer-administered surveys in Spain. Data were collected between April and July 2017, after interviewers were properly trained for the task.
Participants were shown a stimulus in the form of a website of a fictitious company and then responded to the questionnaire.
We focused on the website for two main reasons. First, a website is the most frequent medium used to engage in CSR communication because it provides a highly accessible but inexpensive medium to avoid accusations of spending more on communication than on the initiatives themselves . Second, websites are a preferred medium to communicate CSR involvement because of the richness of argumentation and opportunities for interactivity they provide .
A fictional stimulus was purposely used to control for participants' knowledge, attitudes, and behavioural intentions concerning real companies, therefore avoiding their influence on the model proposed in this study (Kim, 2014).
The website contained information concerning the CSR activities and investments of the company, especially focused on the fight against childhood leukaemia (Supporting Information Appendix S1).
This social cause was chosen based on previous studies that had considered health as a critical issue for CSR assessment (Currás, 2007;Nan & Heo, 2007).
To ensure the variability needed to check the hypotheses in the empirical model, we collected data from two independent samples.
In doing so, we also aimed to control for pre-established attitudes toward business sectors, examining the model in two business sectors (Kim, 2014). In the first sample, we simulated that the website was from an ethical bank (i.e., Your Bank), while in the second sample the website was linked to a chain of ecological restaurants (i.e., Ecofood). Banking companies and restaurant chains often have been compared in literature as they represent the contrast between utilitarian and hedonic products (Andreu et al., 2015).
Instead of exploring "general" companies, we chose an ethical bank and an ecological restaurant to avoid negative biases regarding the motives of companies to collaborate with social causes. In this regard, ethical and ecological companies have CSR at the core of their business.
Therefore, it seems that investing in CSR is a natural fit for them, and this could reduce scepticism and enable participants to focus on evaluating the message content without a negative predisposition towards it.
We used a non-probabilistic sampling procedure to design both research samples. To guarantee a more accurate representation of the data, we used multi-stage sampling by quotas based on participants' age and gender. After data collection and processing, 302 valid surveys remained (response rate = 77.2%), with 150 participants evaluating the bank scenario and 152 participants evaluating the restaurant chain scenario.
It is also important to notice that the content of the website was not manipulated a priori according to the variables in our conceptual model (i.e., the message content was exactly the same in each scenario). Therefore, a questionnaire was administered to participants to openly register the diverse perceptions that each person could have of each dependent and independent variable in the scenarios.
The questionnaire included 14 questions related to the content of the message in the website (issue importance, commitment, impact, motives, fit), external outcomes motivated after reading the message (purchase, advocacy), internal characteristics (issue support), and several classification and demographic traits of the participants (gender, age, education, income).

| Measurement scales
We used a 7-point Likert-type and semantic differential scales to measure the constructs in the model, where 1 represented the participant's total disagreement with the proposed statement and 7 meant total agreement with it.
The 5-item scale (FIT1-FIT5) used to measure CSR fit was adapted from Speed and Thompson (2000) and Skard and Thorbjornsen (2014). CSR commitment was measured by means of a 5-item scale (COMM1-COMM5) taken from Walton (2014). Purchase (PURC1-PURC3) and advocacy (ADVO1-ADVO3) were evaluated with two 3-item scales adapted from the original proposals of Groza, Pronschinske, and Walker (2011) and Romani, Grappi, and Bagozzi (2013), respectively. Finally, issue support, which is one of the two moderating constructs in the study, was measured by means of a 5item scale (SUPP1-SUPP5) adopted from Mittal (1995). All the items are presented in Table 1.
Given that all the measurement scales were originally developed in English and the questionnaire was administered in Spanish, we used a back-translation procedure to check for translation accuracy.
First, we translated the scales from English to Spanish. The new questionnaire was revised and back-translated to English by a proof editor, who guaranteed the conceptual equivalence of the two idiomatic versions of the survey.

| Hypotheses testing
The hypotheses were tested with structural equation modelling (SEM) using the software EQS 6.1. For this purpose, we first implemented a first-order confirmatory factor analysis (CFA) that included all the constructs of the model, taking into consideration the global sample of the study. Second, we implemented the SEM estimation for the global sample, using the robust maximum-likelihood procedure that avoids problems related to non-normality of data by providing a robust chi-square statistic and robust standard errors.
Subsequently, we implemented two multisampling analyses to test the moderating role of issue support and industry in the responses to CSR message content. In the first one, the global sample was segmented according to participants' support for CSR (low vs. high). For this purpose, we implemented the procedure suggested by Bordonaba and Polo (2008). Specifically, we calculated the mean value of the moderating construct and used it as a cut-off value to discard those participants who fell outside the interval determined by the mean ± SDx1/4. In the second analysis, the empirical model

| Confirmatory factor analysis
This section summarises the results of the test of the psychometric properties of the measurement scales used in the study. Tables 2 and   3 show the results of the first-order CFA. To evaluate the quality of all the indicators that are explained in this section, the recommendations of Hair, Black, Babin, and Anderson (2010) were followed. Table 2, the findings confirmed that the Satorra-Bentler chi-square was significant (S-Bχ 2 (430) = 677.40, p <0.01),

Issue importance IMPO1) The company transmits that this is an important cause; IMPO2)
The company transmits it is vital to tackle this cause; IMPO3) The company transmits that companies have a responsibility to address this cause

CSR impact
The information presented in the website is … IMPA1) Abstract/Concrete; IMPA2) Ambiguous/Clear; IMPA3) Not descriptive/Descriptive; IMPA4) Not vivid/Vivid; IMPA5) Not easy to imagine/Easy to imagine

CSR motives
The motivations of the company to support the cause are … MOTI1) Self-interested/Community interested; MOTI2) Firm-focused/ Customer-focused; MOTI3) Profit-motivated/Socially-motivated CSR fit FIT1) The image of the cause and the image of the company are similar; FIT2) The company and the cause fit together well; FIT3) The company and the cause stand for similar things; FIT4) It makes sense to me that the company sponsors this cause; FIT5) There is a logical connection between the cause and the company CSR commitment COMM1) The company seems to feel strongly about helping the cause; COMM2) The company demonstrates a real interest in making an impact to help the cause; COMM3) The company is capable of long-lasting beneficial effects towards the cause; COMM4) The company seems like they will support the cause for a long period of time; COMM5) The company will more than likely make a large impact toward helping the cause value (0.04) was below the maximum limit of 0.08 recommended in literature.
We evaluated the reliability of the measurement scales by means of the Cronbach's alpha (α), composite reliability (CR), and average variance extracted (AVE). Table 2 shows that for all the constructs in the model these indicators were greater than the recommended values of 0.70, 0.70 and 0.50, respectively.
The convergent validity of the scales also was confirmed because the t-statistic revealed that all the items were significant at the confidence level of 95% and their standardised lambda coefficients (λ) were greater than 0.50 (Table 2).
To test the discriminant validity of the measurement scales, we used the procedure suggested by Fornell and Larcker (1981). The results also verified the discriminant validity of the constructs because, when compared in pairs, the AVE estimates of the constructs under scrutiny always exceeded the squared correlation between them (Table 3).

| Multisampling analyses
For the multisampling analyses, first we estimated the empirical model in each subsample of participants segmented according to issue support (low vs. high) and industry (bank vs. restaurant chain).
The standardised lambda coefficients and their t-statistics are presented in Table 5.
The factorial invariance of the model was confirmed by two analyses (p > 0.05 in 100% of the λ compared for issue support and industry), which demonstrated that the measurement model was appropriate for understanding the responses of different types of people to the content of the CSR message.
We finally proceeded to study the structural invariance of the model which enabled us to test research hypotheses H5 and H6. The findings showed that issue support (Dif.S-Bχ 2 (34) = 77.10, p < 0.01) and industry (Dif.S-Bχ 2 (26) = 38.18, p < 0.10) moderated the model significantly. Nonetheless, these two variables only affected some relationships (Table 5). Therefore, the hypotheses H5 and H6 were only partially supported by our findings.

| D ISCUSS I ON AND CON CLUS I ON S
The findings of this study demonstrate that the conceptual model In this regard, CSR and communication researchers need to acknowledge the relationships that have manifested in this study between CSR fit and issue importance, CSR fit and CSR commitment, and CSR motives and CSR commitment. These relationships have not been tested in depth previously and therefore the field would benefit significantly from further analysis of these interconnections.
First, the findings corroborate that four of the five variables related to CSR message content are directly associated with purchase and advocacy responses. This is the case for perceptions on how the message highlights the importance of the social cause (issue importance), the impact the company's support has on the cause (CSR impact), why the company engages in the cause (CSR motives), and the congruity between the cause and the company's core business (CSR fit).
Issue importance relates to purchase and advocacy because it allows for awareness of the cause and its relevance for society, which therefore improves the effective processing of CSR appeals by aligning stakeholders' interests to the cause and company (Auger et al., 2003;Pomering & Johnson, 2009 company's degree of collaboration with a cause is very effective for CSR communication because it enables people to understand the true level of a company's CSR involvement (Pomering & Johnson, 2009).
Third, perceptions of a company's altruistic or intrinsic motivations to collaborate with a cause makes people infer corporate transparency and sincerity, which are highly appreciated when buying products or recommending companies to other people . Fourth, CSR fit shows corporate consistency, therefore improving credibility and brand equity, including advocacy responses (Dawkins, 2004;Lai et al., 2010).
Nonetheless, our findings suggest that perceptions of CSR fit do not have a direct significant impact on purchase. On the contrary, their effect appears to be mediated by issue importance and CSR motives. This finding may be explained by the fact that the company and CSR message explored in this study were fictitious. As explained by Nan and Heo (2007)  Note. T-statistic: *p-value < 0.10; **p-value < 0.05; ***p-value < 0.01.
Electronic copy available at: https://ssrn.com/abstract=3587840 when they are unaware of the company/brand under scrutiny. In our research, participants were unaware of the company because they had never heard of it before. Thus, a very compromising behaviour such as purchase is not manifested in the context of our research because it would require that participants have further knowledge of the company and its attributes.
Second, and contrary to these four aspects of the CSR message content, the company's long-term commitment to the cause (CSR commitment) has no effect on either purchase or advocacy responses. This variable proved to be problematic in previous research. For instance, the findings concerning CSR commitment in the experimental study by Pomering and Johnson (2009)  These findings are explained by the fact that CSR communication has a stronger personal resonance among people who strongly support CSR and therefore it increases their motivation to process the message, which is more effective than neutral messages that do not move them MacInnis et al., 1991).
People with low support of the cause give more importance to the CSR motives of the company when evaluating its long-term commitment to the cause and reporting their purchase intentions. In this regard, CSR motives is the only variable that is given more importance among the low support segment. Therefore, it seems that people who are not supportive of the social cause championed by the company are more distrustful of its CSR communication than highly supportive people.
As far as the role of industry is concerned, the findings suggest that the effect of this variable is not as straightforward as the impact of issue support. More precisely, the empirical model fits both industries relatively well, although numerous differences are observed in the intensity of the associations among the constructs in the two subsamples. For instance, CSR motives prove to be significantly more relevant in the banking scenario because they improve not only purchase but also advocacy responses, which are not improved by CSR motives in the restaurant scenario. These ideas are justified by institutional theory, which suggests that companies in high-profile industries face greater scrutiny from stakeholders, who are especially sceptical and critical of the CSR communication coming from companies (Pérez et al., 2015;Roberts, 1992). For participants in the restaurant scenario, information that relates to issue importance and CSR commitment is more relevant than in the banking context.
The differences between subsamples are especially relevant for issue importance, which is directly associated with purchase and advocacy responses among participants in the restaurant scenario while it does not have any significant impact in banking. This time the finding can be justified with arguments taken from the literature that has defined consumer responses to hedonic versus utilitarian services.
Researchers have demonstrated that, when evaluating a hedonic service (i.e., restaurant chain in our study), consumers primarily use emotional clues to rate the message and the company (Kempf, 1999).
In contrast, consumers evaluate utilitarian products (i.e., bank in our study) on the basis of cognitive criteria (Andreu et al., 2015). In the context of our research, some aspects of CSR message content are more closely connected to cognitive issues (e.g., CSR impact, fit, or motives), while other variables relate to affective aspects (e.g., issue importance, CSR commitment). Thus, issue importance represents an affective assessment that is more significant for people when evaluating the CSR message of a restaurant chain than a banking company.

| LIMITATI ON S AND FUTURE LINE S OF RE S E ARCH
This study is not without limitations, and future research should consider them to improve our knowledge on CSR communication.
First, we used a relatively small convenience sample that limits the generalisation of our findings. The fact that the sample was exclusively collected in Spain also represents a limitation in terms of how the findings of the study should be interpreted and generalised to larger populations. Thus, future studies could benefit from using larger samples collected in different country settings.
Also, the use of fictitious companies and CSR information that was not real can limit the generalisation of our findings to correctly represent support responses in real contexts. If real companies were explored, future studies would need to consider additional moderating variables related to several corporate characteristics, such as prior corporate reputation or CSR positioning (Du et al., 2010).
Finally, a new line of research that has proved to be relevant for management decision relates to the role of gender in Electronic copy available at: https://ssrn.com/abstract=3587840 CSR communication and its effect on purchase behaviour (Kim & Ferguson, 2014). Just as it happens with the presence of women on business boards (Samara, Jamali, & Lapeira, forthcoming), practitioners should acknowledge there is a clear gender difference in communicating CSR. Females tend to accept CSR communication more than males, while they are also more sensitive to CSR messages with self-promotional tone, message transparency, and consistency of CSR communication than males (Kim & Ferguson, 2014).
Therefore, future research should take into consideration gender when exploring customer external reactions to the different elements of the CSR message content.

CO M PLI A N CE WITH E TH I C A L S TA N DA R DS
Author A declares that she has no conflict of interest. Author B declares that she has no conflict of interest. Author C declares that he has no conflict of interest.
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
Informed consent was obtained from all individual participants included in the study.