Determinants of Electronic Word-of-Mouth on Social Networking Sites About Negative News on CSR

Social network sites are a new communication channel to convey CSR information. They are interactive channels that let users participate, spread content and generate positive and negative electronic word-of-mouth about companies (eWOM) that can dramatically affect their reputation and future business. To identify the factors behind this behaviour, we designed a causal model to explain the intention to both comment on and share a negative corporate social responsibility (CSR) news posted on Facebook. We included the following as explanatory variables: social consciousness, environmental consciousness, information usefulness, corporate image and attitude towards sharing and commenting on CSR news. We surveyed 208 Facebook users who evaluated a fictitious news item about an environmental issue. We observed that social and environmental consciousness influences the usefulness of information and the attitude towards behaviour, which may explain the eWOM intention of the specific news. However, corporate image can cause people to refrain from generating eWOM and spreading the news on their friend social network.


3
In the realm of CSR, SNSs can help to communicate these initiatives and increase the effectiveness of communication (Lee et al. 2013;Kesavan et al. 2013). These media favour the engagement, that is, the public interaction with the information, measured this by the number of likes, shares and comments for each post (Albitbol and Lee 2017). With regard to this, Lee et al. (2013) found that social media rewards the socially responsible firms with an advantage in achieving stronger and faster social transmissions by users. However, SNSs are "people's media" (Kesavan et al. 2013, p. 61), and users can freely express positive impressions about organisations, but also negative opinions. In the latter case, they may produce a very damaging form of advertising that can significantly affect a company's reputation and future business (Chan and Ngai 2011;Einwiller and Steilen 2015;Balaji et al. 2016). The current paper analyses this issue.
We are facing a new type of eWOM with considerable influence on the performance and reputation of companies (Chan and Ngai 2011;Einwiller and Steilen 2015;Balaji et al. 2016), but despite its importance, it is poorly documented in the literature (Ho and Dempsey 2010;Chu and Kim 2015;Lee and Ma 2012), particularly regarding CSR information. To contribute in this realm, our research focuses on the factors explaining eWOM on a SNS (Facebook) for a negative CSR news, that is, a piece of news that potentially presents the irresponsible behaviour (environmental dimension) of a company. This paper makes three original contributions to the literature and business practices.
First, much CSR information reach people through mass media, widely perceived as a non-corporate and highly credible communication source to convey information (Du et al. 2010;Eisend and Küster-Rohde 2012). The news media can be an ally in the promotion of CSR (Zhang and Swanson 2006), but journalists will always cover irresponsible actions (Tench et al. 2007;Pérez et al. 2018). In fact, on an almost daily basis the media report small and large cases of irresponsible behaviour (Lin-Hi and Muller 2013). Taking into account that consumers with easy access to information through social media have become more responsive to ethical and sustainability issues (Ballew et al. 2015) and CSR information has a high impact on consumer engagement (Uzunoğlu et al. 2017), it can be expected that SNSs punish the socially irresponsible firms, with a strong viralisation of information (Lee et al. 2013). Although there are works analysing the relationship between news sharing and social media (Berger and Milkman 2012;Kümpel et al. 2015;Olson 2015, 2016), to the best of our knowledge there have been no studies focused on CSR-specific news. Furthermore, analysis of the factors that motivate users to share information and produce eWOM regarding this type of news is still underexplored.
Second, company environmental performance is identified within the broader categories of CSR (Carrington et al. 2010) and is one of the most frequently used dimensions of the CSR construct (Klein and Dawar 2004). There is a widespread public concern for the environment, explaining the growing importance of companies' ecological conduct (Grimmer and Bingham 2013). While a growing number of studies focus on the impact of companies' proenvironmental performance in consumer decision making (Schlegelmilch and Bohlen 1996;Choi and Ng 2011;Grimmer and Bingham 2013;Martinez et al. 2018), few studies focus on reactions to irresponsible behaviours in this regard (environmental degradation or pollution, among others) (Jones et al. 2009). For example, environmental activists have taken the use of oil palm viral to force Nestle to take corrective measures, and Nike also received unfavourable feedback about the irresponsible labour practices of its suppliers, compelling it to amend its business practices (Kesavan et al. 2013). This paper will try to shed light about factors motivating this behaviour.
Third, unlike previous studies, eWOM in social networks is measured as the intention to both share and comment on specific online content. One of the more attractive characteristics of social media is its support for user-generated content, transforming individuals from passive consumers of content to active producers (Nov et al. 2010). According to Kaplan and Haenlein (2010), when users click a 'like' or 'share' button, they are engaging in the replication of already existing content (e.g. posting a copy of an existing newspaper article), and are interacting with-but not generatingcontent. Comments, the highest form of engagement, give users the ability to engage directly in conversation with the company (Cho et al. 2014). We therefore decided to analyse this specific response-sharing and commenting on CSR news-as a contribution to the field.
Finally, this research focuses on Facebook because, according to the statistics, it is the largest (Kesavan et al. 2013) and most successful (See-To and Ho 2014) SNS. Currently, its usage penetration among Internet users in leading markets is above 80% and it has 2.23 billion monthly active users around the world (https ://www.stati sta.com, second quarter 2018). Among companies that use social media, Facebook is one of the most frequently used online platforms and the most valuable social tool for driving engagement with customers (Kesavan et al. 2013).
The remainder of the paper is structured as follows. First, we propose the conceptual model and the hypotheses to be tested in the research. Second, we describe the methodology. Then, we present the results, and we conclude by presenting the most relevant findings, theoretical and management implications, limitations of the study, as well as future lines of research.

Literature Review and Research Hypotheses
Traditionally, the discussion of CSR has predominantly focused on "doing good" and enhancing the welfare of stakeholders (Lin-Hi and Müller 2013), which translates into significant competitive advantages in terms of reputation and consumer loyalty (among other benefits) (García-De los Salmones et al. 2005;Luo and Bhattacharya 2009). However, CSR also refers to avoiding irresponsible acts (Vaaland et al. 2008) or "avoiding bad" (Lin-Hi and Müller 2013). Corporations have long been criticised for irresponsible actions such as pollution, damage to the environment and the unfair treatment of employees and suppliers (Murphy and Schlegelmilch 2013). In these cases, the public disclosure of irresponsible behaviours entails a variety of negative consequences (Sweetin et al. 2013), such as customer losses, decreased employee motivation, reputational damage (Lin-Hi and Müller 2013) and negative word-of-mouth (Grappi et al. 2013).
Consumers appear to be particularly susceptible to a company's CSR initiatives, and it has been found that they are more sensitive and react to a greater extent to negative than positive CSR information (Mohr et al. 2001;Bhattacharya and Sen 2004;Schultz 2011). According to previous research, there is significant asymmetry between the cognitive processing that people do before negative (adverse or threatening) events and the processing they do in response to positive ones (Fiske and Taylor 2013). Specifically, when they are exposed to an irresponsible behaviour, people will spend more time thinking about it and search more extensively for causal information, and accordingly their resulting judgments and responses will be more extreme (Fiske and Taylor 2013).
As has been highlighted previously, one class of consumer reaction is word-of-mouth. It involves sharing with others one's emotions, distaste, disapproval or disparagement concerning irresponsible actions by corporations, and implicates communication and social influence (Grappi et al. 2013). With the Internet and social networks, users can react to information about socially irresponsible corporate behaviours (e.g. by liking, commenting or sharing through their social profiles; see Chu and Kim 2015;Chang et al. 2015), generating eWOM. This opinion-passing behaviour and generation of content facilitates the flow of information, multidirectional communication and the viralisation of content with a few clicks of the mouse (Dellarocas 2003;Norman and Russell 2006;Ho and Dempsey 2010;Lee et al. 2013).
To the best of our knowledge, there have been no studies focused on CSR-specific news. Kümpel et al. (2015) conducted a literature review of papers regarding the relationship between news sharing and social media. However, these studies did not distinguish between different types of news, and failed to consider how individuals hold different preferences that can influence their sharing behaviours (Mitchelstein and Boczkowski 2010;Lee and Ma 2012). Berger and Milkman (2012) investigated the types of articles that are highly shared from the New York Times website, considering content's valence (whether an article is positive or negative) and the extent to which they evoked various specific emotions. Olson (2015, 2016) also analysed the interactions of users with online news, noting that their practices in terms of commenting and passing along media content are connected to their preferences, that is, what they 'fancy' and 'care about' (Tenenboim and Cohen 2015). For example, the authors found that online users do not interact with most news items, as few topics attracted their attention, among them stories related to 'changes in the physical space', that is, news pieces concerning events or courses of events taking place in citizens' everyday environments that seemed to have some type of encumbering effect. CSR activities in general, and environmental issues in particular, could be included in this category. Companies are expected to benefit the natural order or at least minimise any impact, and their actions concerning these issues can make users react. However, the factors behind such behaviour remain unknown.
Considering this gap, we try to contribute to the literature by studying the antecedents of the intention to generate eWOM (share and comment) on one SNS (Facebook) about a negative CSR piece of news, specifically about a potentially irresponsible environmental behaviour. We used the integrative framework of the impact of eWOM communication (Cheung and Thadani 2012) as a starting point to design the causal model. Specifically, we explore how factors related to the receiver (Facebook users) about the stimulus (post content) influence the response (intention to share and comment). Concretely, we include as explanatory variables users' beliefs, attitudes (Chang et al. 2015) and social and environmental consciousness as factors related to information processing (Cheung and Thadani 2012).
The literature about eWOM is rather fragmented and the interrelationships among key variables have yet to be systematically reviewed and studied (Cheung and Thadani 2012). Therefore, in order to support the causal relationships and frame our research, we consider diverse theories of communication. Consumer reactions to CSR news about irresponsible behaviours are dependent upon the individual's attention and interpretation (Lange and Washburn 2012), hence we use human information processing theories such as Elaboration Likelihood Model, commonly used in the study of the impact of eWOM communication (Cheung and Thadani 2012). Furthermore, we use the Value-Beliefs-Norms Theory (Stern et al. 1999;Stern 2000) to justify that between values and behaviour appear other factors, like attitudes.

Social and Environmental Consciousness and eWOM Intention
The personal characteristics of receivers may influence their beliefs and attitudes towards the stimulus (Cheung and Thadani 2012). The impact of information received may vary from person to person, and the same content can provoke different responses in different receivers (Chaung and Thadani 2012). Communication about sustainability should resonate with consumers on the basis of their moral and other foundational attitudes and beliefs (Watkins et al. 2015), so social and environmental consciousness can be key variables of the eWOM intention model.
On the one hand, social consciousness is defined as an individual's concern about the community and its people (Ladhari and Tchetgna 2017). On the other hand, environmental consciousness involves specific psychological factors related to an individual's propensity to engage in proenvironmental behaviours and make green purchasing decisions (Schlegelmilch et al. 1996;Zelezny and Schultz 2000;Martínez et al. 2018). Socially conscious individuals are aware of how they are influenced by others, as well as how they actions may affect others (Schlitz et al. 2010). They are characterised by their understanding, tolerance and appreciation of the well-being of other people, increase in quality of life and economic progress in their communities and protection of nature (Webb et al. 2008;Pepper et al. 2009;Kim et al. 2010). Such individuals behave according to their interests and concern for social issues, with the environmental factor being especially important (Pepper et al. 2009;Collins and Kearins 2019; Pérez and García-De los Salmones 2018).
Thus, social and environmental values are different concepts (De Groot and Steg 2009), but they are related (Kaiser and Byrka 2011). People who give priority to collective, or prosocial, values tend to be more environmentally concerned (Gifford and Nilsson 2014) and have stronger proenvironmental beliefs (De Groot and Steg 2009). In this line, Allen and Ferrand (1999) hypothesised that to act proenvironmentally, individuals must focus beyond themselves and be concerned with the community at large. Prosocial traits would be reflected in people's environmental engagement (Kaiser and Byrka 2011) and lead to proenvironmental behaviours, because it entails that people benefit others, whereas often, no direct individual benefits are received by engaging in these behaviours (De Groot and Steg 2009). We therefore propose that: H1 Social consciousness positively influences environmental consciousness.
A reader's motivation to process the information contained in a message-that is, the person's innate desire to think about such information-can influence their beliefs and attitudes towards opinion-passing behaviour (Gupta and Harris 2010;Cheung and Thadani 2012). This argument is supported by the elaboration likelihood model (ELM; Petty et al. 1983;Petty and Cacioppo 1986), a dual process theory of persuasion that researchers often use for behavioural changes among message receivers (Chang et al. 2015). ELM defines central and peripheral routes based on the type of information processed by a given user, and it explains the circumstances under which that user may be more influenced by one route than the other (Bhattacherjee and Sanford 2006). Depending on a person's motivation, their elaboration likelihood will be either high or low, which will, in turn, determine the route through which persuasion may occur (Petty and Cacioppo 1986). The peripheral route involves little cognitive effort, and the subjects rely on cues regarding the target behaviour. People using central routes, however, evaluate the issue based on the relevant arguments presented to them and carry out more cognitive activity and elaboration (Petty and Cacioppo 1986). The central route processes message-related arguments, requires thoughtful comprehension of the information presented, evaluates the quality of the information and combines multiple arguments into an overall evaluative judgment. Due to this higher cognitive effort, people have to be motivated to engage with the information, and this motivation can fluctuate with situational contexts and time, even for the same individual (Bhattacherjee and Sanford 2006).
In the realm of CSR communication, the segments of stakeholders who are prosocial, CSR advocates or activists will be more motivated to process such information, they will evaluate the message to a higher extent using central routes and the communication will be more effective (Du et al. 2010). If people view a given message (e.g. a post with a piece of news about CSR) as being important and relevant (e.g. due to their social and environmental consciousness), they are more likely to invest the necessary cognitive effort to scrutinise the information content, and they will carefully consider the quality of the arguments presented. Those who view the same message as having little personal relevance, however, may not be willing to spend the time and effort to analyse it (Bhattacherjee and Sanford 2006;Cheung and Thadani 2012). Social and environmental consciousness will therefore enhance the usefulness of the news item, especially when it refers to a potentially irresponsible corporate behaviour. We propose that: H2 Social consciousness positively influences the information usefulness of a negative news item about CSR on Facebook.
H3 Environmental consciousness positively influences the information usefulness of a negative news item about CSR on Facebook.
Social and environmental consciousness may explain attitudes towards participating in social networks, in this case, sharing and commenting on CSR news on SNSs. According to Schlitz et al. (2010), worldview transformation has led to different levels of social consciousness, from passive citizens (embedded level) to people with an active engagement with the world around them, and a strong sense of interconnectedness with others. When a person is aware that their participation in the social environment has an impact (self-reflexive level), they feel motivated to act, taking individual actions (engaged level), collaborative ones, i.e. sharing stories, experiences and ideas (collaborative levels) and/or participating in a field of shared experience in social groups (resonant level). The different levels of consciousness emerge in accordance with what the social environment can reflect, and consciousness expands as long as private experience is expressed publicly (Whitehead 2001). SNSs can boost this evolution and the active participation of people around these issues, because they are public media, and make possible to exchange ideas, interests and create community (Chen 2011;Chiang 2013).
Specifically, previous research has found that social media users have become more responsive to ethical and sustainability issues (Ballew et al. 2015) and CSR information favours consumer engagement (Uzunoğlu et al. 2017). Thus, it can be expected that the more socially consciousness the more predisposition to interact with CSR information in social profiles. With regard to this, the Value-Beliefs-Norms theory (Stern et al. 1999;Stern 2000) postulates that the relationship between values (e.g. social and environmental values) and actual behaviour (e.g. intention to comment and share a specific item on CSR) is mediated by factors other than specific attitudes. According to this theory, an important antecedent to proenvironmental behaviour is the activation of a personal norm, experienced as a moral obligation to act, create a willingness to act proenvironmentally and protect whatever is threatened, derived from the individual's relevant social and environmental values (Jansson et al. 2011;Nordlund and Garvill 2002). Thus, altruistic (altruism towards other humans) and biospheric (altruism towards other species and the environment) values have a positive association with green consumer behaviour (De Groot and Steg 2009), because they may active personal norms. Following this framework, Golob et al. (2019) also found that values that refer concern for the welfare and interests of others (self-transcendent values), active beliefs and personal norms/obligations to make socially responsible purchases.
All these arguments can be extended to the context of our research. Social and environmental consciousness may induce individuals to actively seek information about valued objects and/or act as an amplifier of available information (Stern and Dietz 1994). Thus, these individuals will have more predisposition to participate in the social network sharing/commenting on CSR news, generating higher levels of consciousness thanks to the opportunities of interactivity, dialogue and shared space offered by these media. We therefore propose the following: H4 Social consciousness positively influences attitudes towards sharing and commenting on news about CSR on Facebook.
H5 Environmental consciousness positively influences attitudes towards sharing and commenting on a negative news item about CSR on Facebook.

The Effect of Beliefs and Attitudes on eWOM Intention
Beliefs and attitude influence intention to like and share a post in Facebook (Chang et al. 2015). When users read a post, their beliefs about the usefulness of the content are formed (Chang et al. 2015). Information usefulness refers to its informative qualities and usefulness to the individual (Bhattacherjee and Sanford 2006;Chang et al. 2015;Erkan and Evans 2016), as well as the relevance, accuracy and comprehensiveness of the information available to the consumer (Cheung et al. 2008). We hypothesised previously that highly motivated people to process the information will use central routes, and this higher cognitive effort will be more predictive of subsequent behaviour, as opposed to peripheral routes (Petty and Cacioppo 1983). User perceptions of the value of the information available through eWOM sources have often been identified as a determinant of the influence exerted by these sources on their behaviour (Erkan and Evans 2016;Herrero et al. 2015;Herrero et al. 2018). Specifically, this variable has been found to have a positive relationship with eWOM adoption (Cheung and Thadani 2012) and opinion-passing behaviour (Chang et al. 2015). On SNSs, people encounter a great quantity of information, so they might have a greater intention to adopt or react towards this information only so long as they find it useful and interesting (Erkan and Evans 2016;Almgren and Olsson 2016). In fact, it has been found that content that is considered interesting or contains high informational value factors such as surprise, controversy, relevance or unexpectedness, has a positive relation with sharing (Berger and Milkman 2012;Rudat et al. 2014;Kümpel et al. 2015). Considering these arguments we propose that: H6 Information usefulness positively influences the intention to share and comment on negative news item about CSR on Facebook.
When the negative information is related to a company, corporate image can play a role in eWOM intention. Previous research has found that a good corporate image-defined as an intangible asset resulting from the interaction of all the experiences, beliefs, feelings, knowledge and impressions that people have about a firm (Berstein 1984)-provides a number of benefits to firms (Rhee and Haunschild 2006), and it can be an important resource during a crisis (Beonit 1997;Dean 2004;Klein and Dawar 2004;Coombs and Holladay 2006;Insch and Black 2018). Consumers spontaneously construct beliefs and attributions of blame for faulty or harmful products (Folkes and Kotsos 1986) and these form the basis of consumers' brand judgments and behaviour. In constructing these attributions, consumers rely on information, including corporate associations (Folkes et al. 1987) and prior corporate image (Dean 2004), which may have a halo effect, as benefit of the doubt and/or a shield (Klein and Dawar 2004;Coombs and Holladay 2006;Insch and Black 2018). If a stakeholder holds a generally favourable view of the organisation, he or she might give the organisation the 'benefit of the doubt' by assigning the organisation less responsibility for a crisis than would be assigned to an organisation with an unknown or unfavourable reputation (Fombrun 1996); this will result in less damage to the organisation's reputation (Coombs and Holladay 2006;Klein and Dawar 2004). Offering a different perspective, Rhee and Haunschild (2006) explore the possibility that companies with a good reputation suffer a more marked penalty in certain circumstances. Specifically, they identified a negative effect of a high-quality corporate reputation on market responses to product defects in the automobile industry. Therefore, it would appear that a good reputation does not necessarily buffer firms from the "vagaries" of the market (Rhee and Haunschild 2006, p. 114). However, it is important to note that the study only focused on the quality dimension of reputation. As the authors recognise, this can be a limitation of the study, as there are many other dimensions underlying organisational reputation. We adopt the perspective of previous articles that measured corporate reputation in a global way (Coombs and Holladay 2006) or considered social responsibility reputation (Klein and Dawar 2004;Insch and Black 2018), and that found a positive effect of corporate image, extending the argument to the area under study. Concretely, we expect that, when people are exposed to a negative news item about a company, they can refrain from sharing and comment it-so long as the company has a prior good image-and wait to see what will happen next or to find more information about the issue. We explore this argument and propose that: H7 Corporate image negatively influences the intention to share and comment on a negative news item about CSR on Facebook Finally, the relationship between attitude and intention has been well established and validated in consumer behaviour research (Cheung and Thadani 2012). In the online context, several studies have used the theory of reasoned action (TRA; Fishbein and Ajzen 1975) and the theory of planned behaviour (TPB; Ajzen 1991) to link attitude and intention. For example, the attitude towards a behaviour explains consumers' engagement in eWOM on Facebook and/or Twitter (Wolney and Mulleer 2013), individuals' intention to participate in viral marketing dynamics (Camarero and San José, 2011) and SNS use behaviour (Currás- Pérez et al. 2014). In the study context, attitude refers to an individual's predisposition to comment on and share news about CSR on his or her profile on Facebook. Based on the postulates of the TRA and TPB and previous studies we propose that:

Methodology
We conducted a quantitative study based on intervieweradministered surveys in Spain. Data were collected between April and June 2018, after interviewers were properly trained for the task. The empirical study focused on the tourism sector and, more concretely, the news item was about a hotel. Generally speaking, the tourism industry is strongly affected by eWOM and, within it, hotels are probably the most affected (Serra-Cantallops and Salvi 2014). People are used to creating or using eWOM about their touristic activities, so it is interesting to analyse this area with a different focus.
Participants (Facebook users) were shown a stimulus in the form of a fictitious post from a newspaper (i.e. Diario Montañés) related to a negative piece of news about the CSR (environmental dimension) of a hotel (i.e. NH Hotels) and then responded to the questionnaire. Specifically, we elaborated a fictitious but realistic news item about an event in the city with consequences for the environment. The post provided information about the opening project of a new hotel for the chain in the city, which would have a possible environmental impact. According to the news item, the ecologists' associations had sounded the alarm (Appendix 1).
We analysed the antecedents of eWOM intention in a real context, so both the source and the hotel were well known to users. Diario Montañés is a regional printed newspaper, with an online version, and is the leader in the region where the empirical study was carried out, with almost 90,000 followers on Facebook. NH Hotels is a Spanish hotel chain (with one hotel located in the city where the empirical study was carried out) with strong CSR engagement, especially in the dimensions of environmental responsibility and sustainability. Energy, water, recycling and climate change are its main performance areas, and it conveys information about its initiatives in CSR reports and on its corporate website on a regular basis. At the end of the survey, we expressed our appreciation for the respondents' participation and indicated that the content of the post was fictitious, specifically created for research purposes.
The questionnaire used in this study included three-item measurement scales for all of the dependent and independent variables included in the research model. In particular, we used seven-point Likert scales, rating from strong disagreement to strong agreement with the statements presented. Intentions towards sharing and commenting on the news item were measured following Herrero et al. (2017). Three items based on Herrero et al. (2015) were used to measure information value. To assess the attitude towards sharing and commenting, a scale based on the studies by Venkatesh et al. (2003) and Herrero and San Martin (2012) were used. Three items based on Schlegelmilch and Bohlen (1996) were used to measure environmental consciousness. The measurement instrument for social consciousness was taken from Walker and Kent (2013). Finally, corporate image was measured with three items based on Lafferty et al. (2002). Appendix 2 shows the scales used for each factor.
We used a non-probabilistic sampling procedure to design the sample. The survey universe is composed of Facebook users, so in order to guarantee a more accurate representation of the data, we consulted the Annual Study of Social Networks 2017, elaborated by IAB Spain (Interactive Advertising Bureau), to ascertain the demographic profile of users. The percentages by gender and age of Facebook users let us define quotas, and the respondents were randomly selected respecting such quotas. Following data collection and processing, the final sample included 208 valid surveys (see Table 1 with the profile of the sample).

Results
A covariance-based structural equations model (CB-SEM) approach (software EQS 6.1.) was used to test the research hypotheses, using a robust maximum-likelihood estimation procedure. This method avoids the problems related to non-normality of data by providing the outputs 'robust chi-square statistic' and 'robust standard errors', which have been corrected for non-normality (Byrne 2006) and which consequently guarantees the validity of the model estimation. First, the measurement model was estimated with confirmatory factor analysis (CFA) to test the psychometric properties of the measurement scales (reliability and validity). Next, the structural model was estimated to contrast the direct causal effects established in the research hypotheses.

Estimation of the Measurement Model
The results summarised in Table 2 confirm that the factorial model fits the empirical data, as the Bentler-Bonett normed fit index (BBNFI), Bentler-Bonett non-normed fit index (BBNNFI), incremental fit index (IFI) and comparative fit index (CFI) clearly exceeded the recommended minimum value of 0.9. Similarly, the root mean square error of approximation (RMSEA) was within the recommended limit of 0.08, and normed χ 2 takes a value clearly under the recommended value of 3.0 (Hair et al. 2010). The reliability of the measurement scales was evaluated using Cronbach's alpha, compound reliability and AVE coefficients (Bagozzi and Yi 1988). The values of these statistics were, in all cases, above the required minimum values of 0.7 and 0.5, respectively (Hair et al. 2010), supporting the inner reliability of the constructs (Table 1). The convergent validity of the scales was also confirmed, because all items were significant to a confidence level of 95% and their standardised lambda coefficients were higher than 0.5 (Steenkamp and Van Trijp 1991).
The discriminant validity of the scales was tested following the procedure proposed by Fornell and Larcker (1981), which requires the comparison of the variance extracted for each pair of constructs (AVE coefficient) with the squared correlation estimate between these two constructs (Table 3). In all cases, the variances extracted for each construct were greater than the squared correlation between them, which supports the discriminant validity of the measurement scales.

Estimation of Hypothesised Structural Model
Once the psychometric properties of the scales were tested, the model was estimated using robust maximum likelihood. A first estimation of the structural model showed that the consumer's environmental consciousness did not have any significant effect on the perceived usefulness of the information (hypothesis H3 is not supported). Consequently, and following the model development procedure suggested by Hair et al. (2010), we reformulated the research model and eliminated the non-significant relationship. Figure 2 summarises the results for the estimation of the proposed research model, indicating the goodness-of-fit indices of the structural model, the R 2 statistics for each dependent variable and the standardised coefficients and significance level (p value) for each relationship.
The goodness-of-fit indices supported the correct definition of the structural model (Normed χ 2 = 2.34; BBNFI = 0.92; BBNNFI = 0.94; CFI = 0.95; IFI = 0.95; RMSEA = 0.08). According to the empirical evidence obtained, social consciousness was a direct determinant of environmental consciousness (H1 is supported), perception of information usefulness (H2 is supported) and attitudes towards sharing and commenting (H4 is supported). Finally, environmental consciousness exerted a direct and positive influence on attitudes towards sharing and commenting about CSR news on Facebook (H4 is supported), but not on perceived information usefulness (H5 is not supported).
The direct antecedents of intention to share and comment on posts about hotels' CSR on Facebook are the perceived usefulness of the information contained in the post (H6 is supported) and the individual attitude towards sharing and commenting on information about these issues (H8 is supported). In addition, corporate image had a direct and negative effect on the intention to share and comment on negative posts about hotels' CSR on Facebook (H7 is supported). Corporate image therefore mitigates the potential spread of negative eWOM about the company, at least at an initial stage. In conclusion, and according to these results, both social and environmental consciousness have a significant influence on consumers' intention to share and comment on posts about hotels' CSR on Facebook, but this is mediated by other variables such as attitudes towards sharing and commenting about CSR or perceived information usefulness.

Discussion
People are increasingly concerned with issues of CSR responsibility, and are exposed to different channels to receive information on this subject. Typical communication sources are one-way channels, such as advertising, CSR reports or media coverage (Du et al. 2010). However, there is an emerging channel, social networks, that has altered the way in which people and companies share information and communicate with each other. Social networks enable users to become integrated in the communication, creation and exchange of content regarding multiple subjects, including their consumption experiences and other relevant information that fits or reflects their personal values. Today, users can support a company's communication or dramatically affect its reputation and future business with only one click of the mouse.
Previous literature has shown that consumers react quickly to negative CSR information (Mohr et al. 2001). In fact, consumers are more sensitive to irresponsible than to responsible corporate behaviour, with 'doing bad' being more impactful than 'doing good'. Hence managers need to be cognisant of the hazards of being perceived as socially irresponsible (Bhattacharya and Sen 2004). With the power of the Internet the risks are even higher, because if people react to any negative information through their social networks, increasingly significant effects in terms of message diffusion and virality will result. We found that social and environmental consciousness are key variables explaining the intention to share and comment on a negative CSR news item. Social consciousness is one of the most important variables of the intention model due to its positive influence on perceived information usefulness and attitudes towards sharing and commenting. Socially conscious people are concerned about and committed to society, so they are attentive to information that affects the community where they live. A news item with information about the potentially irresponsible behaviour of a company will attract their attention, and they will consider it useful, valuable and important. This variable, information usefulness, has a direct impact on the intention to share and comment on such a news item. Furthermore, prosocial people also have a positive predisposition to share and comment on CSR news on their social profiles, and this attitude towards behaviour influences their intention to spread the news.
We also found that social consciousness and environmental consciousness were closely related. We found that social consciousness leads to greater environmental consciousness, and these values also influence attitudes towards sharing and commenting on CSR information on SNSs. It is striking, however, that the relationship between environmental consciousness and information usefulness was not significant. One reason may be in the presentation of the stimulus. In order to be as realistic as possible, we designed a post similar to a real one, providing little information at first, just enough to understand the content, but without excessive detail (users would have to click to read the entire information). It is possible that environmentally concerned people require more information to evaluate the magnitude of the event, which can explain the lack of significance. Social consciousness is a broader and less specific variable, pertaining to concern about anything related to the well-being of the community as a whole. The content of the post refers to a possible alteration of the natural order, and so although brief it can prove sufficient for prosocial users to consider it valuable and useful. It would be interesting to deepen this issue with other types of messages.
Finally, when people are exposed to a piece of news about a possible irresponsible corporate behaviour, not directly related to service failures but for example to environmental issues, the corporate image inhibits them from sharing and commenting on the news, at least in the beginning. People are aware that with this behaviour they are spreading information that could damage the reputation of the company, and the corporate image seems to act as a shield until they obtain more information related to the event.

Theoretical Implications
The main goal of this paper is identify the variables that would explain eWOM (intention to share and comment) on Facebook with regard to negative CSR information, that is, a piece of news about a potentially irresponsible environmental behaviour. In this regard, although previous research has analysed possible consequences of irresponsible behaviours (Sweetin et al. 2013;Lin-Hi and Müller 2013), reactions on the social media are underexplored. This scarcity of previous research highlights the contribution of our paper.
With regard to social and environmental consciousness, some studies have focused on one of the two variables to explain proenvironmental or socially responsible behaviour (Zelezny and Schultz 2000;Martínez et al. 2018) or they have integrated social and environmental issues into only one construct (Kim et al. 2010;Pérez and García-De los Salmones 2018). The current study considers both types of values separately in order to develop a better understanding of their role in the intention model. In fact, the empirical evidence obtained supports that social consciousness explains environment consciousness, in line with previous works that postulate that both variables are related (Kaiser and Byrka 2011;Gifford and Nilsson 2014).
According to Du et al. (2010), social and environmental consciousness can positively influence the effectiveness of CSR communication, because prosocial people will be more motivated to process the information and will evaluate the message to a greater extent, using central routes. This theoretical argument was elaborated considering positive information and controlled communication, but it can be extended to negative information, publicity and the generation of eWOM. Social media users have become more responsive to ethical and sustainability issues (Ballew et al. 2015), and we have found that social and environmental values are behind these behaviours, in the sense that predispose people to commenting/sharing CSR information. Actually, social networks can expand the levels of social consciousness (Schlitz et al. 2010), since they are platforms for relationship and dialogue, where people can pass from passivity to an active engagement.
CSR information have a high impact on consumer engagement (Uzunoğlu et al. 2017), and social networks can reward socially responsible companies, or punish "bad" companies (Lee et al. 2013). With regard to this, in certain circumstances corporate reputation does not provide protection in a crisis (e.g. the failures of a product offered by companies with a good reputation) (Rhee and Haunschild 2006), but in other cases, it has been proved that corporate image can act as a halo against negative events (Klein and Dawar 2004;Coombs and Holladay 2006;Insch and Black 2018). This finding can be extended to the viralisation of information and generation of negative eWOM.

Managerial Implications
Media, as independent communication channels, fulfil their function of informing society about the issues that may affect it, and so will tend to cover companies' irresponsible initiatives through offline and online editions. If this type of news causes users to react, the media will cover these issues to an even greater extent. Therefore, companies need to enhance their relationship with media organisations as much as possible.
The way in which a third party (e.g. the media) presents, filters and frames CSR information can influence the effect of the information on the audience (Lange and Washburn 2012), and hence public relations practices are always necessary. Companies should avoid controversial events that might be covered with a negative tone by media, and continuously monitor the information published, including social networks. SNSs are distinctive in letting companies know how many people react to the news and in what ways. They provide important information regarding the "feeling" of people before certain events and circumstances.
CSR includes many areas and dimensions, so although companies try to minimise negative events, in the day-today running of the business certain problems or situations that generate media attention may arise. The audience, however, decides whether to share or comment on a negative news item about a company and fosters the spread of this information. This behaviour can be mitigated if the company has a good image. Corporate image can act as a guarantee at first, and depending on the problem, it can cushion any possible crisis of communication or image in social networks by restraining the user from sharing the information, at least to a certain extent and for a certain period of time. However, image does not provide complete and continuous protection. Companies have to know how to respond to any negative information with transparency, honesty and empathy. They have to focus not only on media and journalists, but also on people, and so they must change their language and adapt to this new social landscape. The quick reaction or response of the company and an appropriate and timely communication strategy will be key for overcoming the situation.

Limitations and Future Research
This paper has several limitations that open up some future lines of research. First, the findings are limited to the context analysed. We used a well-known company (hotel) and a negative news item related to environment. Featuring other companies with different reputations and other CSR activities would extend the scope of the findings. Furthermore, it would be interesting to compare negative news with positive information, to analyse whether the audience reacts to a greater or lesser extent to the information depending on the tone of the news. The model could also be complete with new variables related to the transmitter (e.g. credibility of the source) and the communication channel (credibility of the social network or subjective norm, among others).
From a statistical perspective, the use of the SEM approach can also be considered a limitation of this study, given the controversy that exists regarding the effective capability of statistical method to test causal effects. Thus, whereas different authors have highlighted the usefulness and value of SEM for causal modelling (Byrne 2006;Barret, 2007;Martínez-López et al. 2013), the potential of statistical analyses based on correlations to establish causality has also been questioned. Therefore, and despite the fact that SEM is commonly used to empirically test theoretical models with a causal structure (Hair et al. 2010;Martínez-López et al. 2013), it would be interesting for future research to analyse the causal effects observed in this study through other techniques, such as an experimental design, with greater validity to test causality.

Compliance with Ethical Standards
Ethical Approval This article does not contain any studies with human participants or animals performed by any of the authors.
Environmental consciousness (Schlegelmilch and Bohlen 1996) ENVC1 I consider myself to be a person sensitive to environmental problems ENVC2 I think it is important to protect the environment ENVC3 I am concerned about my impact on the environment