eWOM

The first studies on eWOM were presented by (^{Wu, Mahajan, & Balasubramanian, 2003}) about the motivations of people to participate in the internet with reviews or ratings on products, services or companies and proposed three motivating factors (focus-related utility, consumption utility and approval utility). ^{Hennig-Thurau et al. (2004}) extended the motives behind eWOM to eight: venting negative feelings, concern for other consumers, selfenhancement, seeking advice, social benefits, economic incentives, platform assistance and helping the company. In this same line, ^{Goldsmith & Horowitz (2006)} posit that eWOM is more relevant than advertising since it is free of the social prejudices inherent to convenience answers, and define eight motives for consumers to use this information (to reduce their risk, because others do it, to secure lower prices, to get information easily, by accident/unplanned, because it is cool, because they are stimulated by off-line inputs such as TV and to get prepurchase). In the last years, the rapid advance of social networking sites, as the leaders within the Social Media environment, has lead to a reorientation of the studies on eWOM information towards the study of its dynamics in this type of networks. Authors such as ^{Chu & Kim (2011)} have proposed a set of variables based on the social theory that can have a positive effect on the development of eWOM (homophily, trust, normative influence and informational influence), such as the notion that those individuals with a higher number of social ties can establish contact with a higher number of people, which increases the possibilities of influencing others and, therefore, the strength of ties makes eWOM become a far-reaching tool (Rejón-Guardia et al., 2013). Specifically in social networking sites, eWOM information has proven to have direct effects on brand satisfaction, brand loyalty and brand image (^{Balaji, Khong, & Chong, 2016}). Finally, studies on eWOM emphasize the influence on the final purchase decision for a product, with this having a direct effect on questions such as the perceived value of the product (^{Gruen, Osmonbekov, & Czaplewski, 2006}) and purchase intention (^{Matute Vallejo, Polo Redondo, & Utrillas Acerete, 2015}; ^{Salo & Frank, 2015}; ^{Wang, Yu, & Wei, 2012}), Nevertheless, the most recent studies examine how the acceptance of eWOM information affects both the attitude and the purchase decision (^{Matute-Vallejo et al., 2015}; ^{Erkan & Evans, 2016}).

eWOM Information Acceptation Model (IACM) for the purchase

^{Erkan & Evans (2016)} posit that the influence of eWOM on the purchase or repurchase decision requires the sum of two variables: the characteristics of the information and the attitude of consumers that adopt such information. Information from eWOM is found in different formal and informal virtual channels (blogs, social networking sites, forums or virtual communities), and is shared in different ways (reviews on brands, comments, messages between users or simple thoughts) where consumers can provide information to other consumers by means of online consumer reviews of products which can affect the buying decision (^{Matute-Vallejo et al., 2015}). Consumers assess this information to determine its usefulness in the purchase decisionmaking process, repurchase decision-making or to strengthen the loyalty bonds with a brand (^{Erkan & Evans, 2016}). Previous studies have used several models and theories to examine consumers’ information acceptance. Regarding eWOM, the firs studies analysed the transfer of basic information between senders and receivers (^{Bansal & Voyer, 2000}). ^{Cheung et al. (2008)} pointed out that the influence of the information might change from person to person and that the same content can evoke different notions among receivers. Based on these preliminary studies, and adopting behavioural theories, authors such as ^{Sussman & Siegal (2003)} or ^{Shen, Cheung, & Lee (2013)}, propose to integrate this knowledge with the Dual Process Theory, which posits that people can be affected by a message in two routes, which are central and peripheral. The central route refers to the essence of arguments within the message, while the peripheral route refers to issues which are indirectly related to the essence of the message (^{Petty & Cacioppo, 1988}; ^{Cheung et al., 2008}; ^{Shu & Scott, 2014}). From these contributions, the Information Acceptance Model (IAM), which is based on the theory of Technology Acceptation Model (TAM) proposed by ^{Davis (1989)}, was developed. Therefore the model includes four components: argument quality (which represents the central route), source credibility (which represents the peripheral route), information usefulness and information acceptance (^{Sussman & Siegal, 2003}), as illustrated in Figure 1. With this integration, the IAM offers an explanation as to how people are affected by the information on computer mediated communication platforms. IAM models applied to the eWOM are made of three variables (quality of the information, credibility and usefulness of the information), which have direct effects on consumers’ information acceptance (^{Cheung et al., 2008}; ^{Chu & Kim, 2011}; ^{Shu & Scott, 2014}). In particular, ^{Cheung et al. (2008)} has applied this model within the online discussion forums context, while it was applied by ^{Chu & Kim (2011)} within the social media context.

Source: ^{Sussman & Siegal (2003)}.

Figure 1 Information acceptance model. 

Nonetheless, although IAM is a commonly used model in eWOM, it only focuses on the characteristics of information. The influence of information, however, should not be limited to the characteristics of information, consumer behaviour towards information should be evaluated together with them, according to the model developed by ^{Erkan & Evans (2016)} and to the theoretical framework proposed by Knoll (2016), who add consumers’ attitude towards information, as illustrated in Figure 2.

Source: Adapted of ^{Erkan & Evans (2016)}.

Figure 2 Information aceptance model of the eWOM (IACM). 

Information acceptance and purchase intention

Recent studies have shown that there is a significant relationship between the acceptance of eWOM information and the use of such information for the purchase and repurchase of products and services, in social networking sites (^{Wang et al., 2012}), vendor’s websites (^{Matute Vallejo et al., 2015}) or social media in general (^{Erkan & Evans, 2016}). This relationship is basically supported by all the technological adoption models applied to electronic purchases therefore (^{Davis, 1989}; ^{Fishbein & Ajzen, 1975}; ^{Venkatesh, Morris, Davis, & Davis, 2003}; Venkatesh, Thong, Xu, 2012) the following hypothesis is proposed:

Usefulness and information acceptance

The technology acceptance model (TAM), proposed by ^{Davis (1989)}, ^{Davis, Bagozzi & Warshaw (1992)}, and the empirical applications of the model (Cheung et al., 2009; ^{Venkatesh et al., 2003}; ^{Venkatesh et al., 2012}) posit a positive relationship between the perception of technology usefulness and technology use, which has been validated by ^{Erkan & Evans (2016)} and ^{Sussman & Siegal, (2003)} for eWOM information from social media and others. Therefore, the following hypothesis is proposed:

Quality and credibility of eWOM information

Research works such as those carried out by ^{Petty & Cacioppo, (1988)}, ^{Park et al. (2007)} and Lee & Shin (2014) postulate that the quality of online reviews has a positive influence on purchase intention, and consequently the quality of eWOM information can be one of the determinant factors of consumers’ purchase intentions as it is proposed by ^{Erkan & Evans (2016)}. Therefore, the following hypothesis is introduced:

According to (^{Wathen & Burkell, 2002}), information credibility is the initial factor in the individuals’ persuasion process, and it is a previous factor of the perceived usefulness according to the IAM. Therefore, the following hypothesis is proposed:

Measurement tool

In order to contribute to the study into the adaptation of eWOM information in Spain, this research intends to adapt the measurement scales used in the literature. Taking as a reference the scales used in the IACM, a total of 24 items to measure the 8 constructs have been adopted.

Specifically, three items were used for each one of the following constructs: attitude towards information (^{Park et al., 2007}; ^{Erkan & Evans, 2016}), quality, need, information usefulness and information acceptance (^{Prendergast et al., 2010}; ^{Chu & Kim, 2011}; ^{Erkan & Evans, 2016}). For credibility of information, four measurement items were used (^{Prendergast et al., 2010}; ^{Erkan & Evans, 2016}), three items for purchase intention and two for purchase, based on the scales used by ^{Park et al. (2007)} and ^{Erkan & Evans (2016)}. The complete measurement scales can be found in Appendix 1.

The information was collected through an online questionnaire with seven-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree), following the guidelines recommended by ^{Churchill & Iacobucci (2004)} to effectively measure variables that are not directly observable.

Sample and data collection

Due to the fact that this is a study on users of social networking sites, the following target population was chosen: males and females from 20 to 40 years old, users of internet and social networking sites in Spain, at least 3 times a week. The data collection method chosen was online survey, for this being an information collection tool that has already been validated in other studies for this type of research (^{Escobar-Rodríguez & Carvajal-Trujillo, 2014}; ^{Agudo-Peregrina, 2014}). A total of 279 questionnaires were collected; eight of them contained consistency problems in the answers, and therefore the final sample was made of 271 valid questionnaires collected in the period April-May 2016 (Appendix 2).

Analysis of the data and results

In order to estimate the structural equation model proposed, a regression analysis of latent variables was carried through the technique of Partial Least Squares (PLS), using the Smartplus PLS program. This multivariate technique to test structural models is recommended in models of novel character and for small samples (^{Escobar-Rodríguez & Carvajal-Trujillo, 2014}; ^{Matute Vallejo et al., 2015}).

The data analysis was carried out in three phases: in the first phase the measurement model was tested, in the second phase the validation of the structural model was examined, while the third one analysed, by using a multi-group test, whether there are differences between males and females in the relationships included in the model.

Validation of the measurement model

The first step is to determine the convergent and discriminant validity of the constructs and the reliability of each item. The convergent validity of each construct is acceptable, since all them have loadings above 0.5 in the convergent validity test (^{Hair et al., 2014}). The individual reliability of each item is measured by the correlations of the loadings of each item versus each variable. In Table 1, it can be verified that the loadings for each indicator are significant (t >1.96), with all them being validated (Table 1). Additionally, in order to test the internal measurement coherence of all the indicators in relation to their corresponding variables, the Dillon-Goldstein’s test, known as the composite reliability coefficient, is applied, where all the values are above the minimum acceptable of 0.7 (^{Gefen et al., 2000}). Then the Cronbach’s alpha test is applied and all the values above 0.7 are obtained, complying with the minimum values (^{Churchill & Iacobucci, 2004}). Finally, the convergent validity is tested again, taking into account the variance, that is to say, that there is a similar variance between the indicators and its construct, which implies that the average variance extracted must exceed the cut offpoint of 0.5 explained by the indicators (^{Fornell & Larcker, 1981}), as can be seen in Table 2.

The discriminant validity is tested through the comparison between the average variance extracted (AVE) of each variable with the squared correlation between the variables, with all the values obtained for the main diagonal (AVE) exceeding the squared correlations. The matrix shows that each variable relates more intensely to its own items than to the items of the other variables since, the values below the diagonal are smaller than the AVE (^{Fornell & Larcker, 1981}), as shown in Table 3.

Therefore, it can be considered that each variable relates more intensely to its own items than to the items of the other variables (^{Fornell & Larcker, 1981}), as shown in Table 3.

Validation of the structural model

Following with the structural model, a resampling is carried out by the bootstrapping technique, using 1,000 sub-samples extracted from the same research data in order to perform a significance contrast of the parameters included in the model, on the basis that the model has fulfilled its predictive power since the R-square values exceed 0.1 (^{Hair et al., 2014}). The results from the bootstrapping test support the relationships proposed, with all hypotheses being significant, except for the relationship between the variable quality of information and usefulness of information. Therefore all hypotheses are supported, except for hypothesis H3, which is not significant, as can be seen in Table 4 and Figure 4.

*p≤0.05; N.S.= No signifcant

Source: by authors

Figure 4 Validated model without the moderator effect of gender. 

Multigroup Analysis of the moderating effect of the variable gender

Finally, in order to confirm hypothesis H9, a multigroup analysis (EMG) is conducted by examining the possible differences in the eWOM information acceptance model between males and females. The results demonstrate that there are significant differences in three of the relationships. The first one, in the relationship between need of information and information usefulness, based based on hypothesis H5, with B=0.260 for females and B=0.155 for females. The second one, in the relationship between credibility of information and information usefulness, based based on hypothesis H4, with B=0.140 for females and B=0.227 for males (Figure 6). In the third relationship with significant differences, between attitude towards information and purchase intention, based on hypothesis H7, the level of affectation for males is higher, with B=0.303 versus B=0.221 for females (Figure 5-6). In the light of these results, the moderating effect can be supported for at least these three relationships of the model proposed (Table 5-6).

*p<0.05

Source: by authors.

Figure 5 Validated model with the moderator effect of gender (Women). 

*p<0.05; N.S.= No significant

Source: by authors.

Figure 6 Validated model with the moderator effect of gender (Men).