Brand awareness is the actual presence of a brand in the minds of consumers. Biswas (1992) argued that brand awareness refers to the strong impression or link of a brand in the memory of a consumer, as well as the experiences associated with the brand in the consumer’s memory. According to the definition of Aaker (1991), brand awareness is the ability of consumers to identify or recall a brand that belongs to a class of products. Aaker offered that well-known brands are usually what consumers consider during consumption, and so for them, brand awareness is an important part of decision-making assessment in any purchase.

Macdonald and Sharp (2000) presented that brand awareness plays an important role in purchase behaviour because consumers tend to purchase familiar or well-known products based on their experience. A brand with high awareness has the following advantages: firstly, chances are greater that the brand is included in the considered options; secondly, it affects the choice of a consumer over multiple brands; and thirdly, it impacts the selection process while also influencing the formation and intensity of brand association to beautify the brand image (Keller, 2008).

Brand awareness refers to the strength of a brand in the minds of consumers. The more frequently consumers see, hear and are exposed to a brand, the more they remember the brand. Therefore, repeated exposure can increase brand identification so as to improve brand awareness. To enhance brand recall, product- or brand-related links should be created in the memory of consumers. For example, brand spokespersons and impressive advertising can deepen brand recall, thereby enhancing brand awareness (Aaker, 1996). In other words, brands need to be recognised; brand recognition covers familiarity and connectivity; in addition, brands need to be easily recalled and grow synchronously with recognition; otherwise, brands are only known or only thought of, instead of exhibiting high awareness.

Reputation predicted brand awareness, preference for the university and opinion towards the university brand. The teaching quality of the university was the most important factor for enhancing the reputation of the university and its brand. Thus, there was a clear difference and relationship between reputation and brand awareness in higher education and how this influenced students’ decisions. This knowledge has useful implications for management practices in higher education (Brewer & Zhao, 2010).

Because of intense competition, universities have encountered severe challenges in recruiting new students (Bock, Poole, & Joseph, 2014; Joseph, Mullen, & Spake, 2012). Student recruitment is only the start of a long-term relationship that universities should develop with their students. This relationship is not limited to the period that students attend courses but continues after graduation. How universities manage their relationships with students and how students perceive university brand are likely to affect students’ recognition of their university, subsequently deciding their willingness to interact with the university in the future. The definition of brand awareness can be summarised as a brand name that is known to consumers and the ability to clearly identify the brand and to recall the brand.

The challenges faced by universities are becoming more and more complicated nowadays, as the operation and management of school images are not easy tasks. With the increase of market pressure, there is more and more evidence showing that the management awareness of colleges and universities is gradually shifting to a corporate operation and management pattern (Davis, Rensburg, & Venter, 2016).

Kotler and Gertner (2002) maintained that customers develop brand beliefs for each product based on each attribute, and that a combination of beliefs held towards a particular brand constitutes a brand image. Romaniuk and Sharp (2003) discussed that brand image is defined as customer experience, market information, WOM and other sources of brand perception; consumers must be exposed to a variety of market information, so that they can have greater association and perception combinations to form a strong and unique brand image in their mind. Hawkins, Best and Coney (2005) proposed the relationship between consumers and product or service images in which consumers choose product images consistent with their own images; when the relationship between self-images and product brand images is satisfactory to consumers, their product identity will be enhanced.

In an increasingly complex and competitive market, universities have formed sustainable development strategies by adopting branding as a solution. Because the services that universities provide are unique, university branding decisions are a worthwhile topic for discussion. According to Pinar, Trapp, Girard, and Boyt (2014), several dimensions concerning brand equity were crucial for developing a strong university brand. Specifically, regarding the core dimensions of university brand equity, perceived quality of teachers was the most critical dimension, followed by university reputation, emotional environment, brand loyalty and brand awareness.

Nowadays, competition among schools is severe, and their brand image is gradually being valued by different circles, as such an image is considered to be an intangible asset. Yeh (2007) stated that school brand image means the schools establish their names and operational characteristics according to educational objectives and the concept of student services to attract students; thus, image is the sum of all tangible and intangible assets. Tsai (2009) suggested that good brands should construct powerful brand effectiveness to create and retain relationships between customers and brands. Likewise, to acquire students’ identity, schools should not neglect the establishment and operation of brand image (Wu & Chen, 2012).

Statements by previous scholars demonstrate the importance of a positive school brand image. Regarding internal variables, a positive school brand image can enhance cohesion to schools, attract excellent talents, form a positive learning atmosphere and reinforce instructional quality. Regarding the external variables, a positive school brand image can be recognised by academic circles and society, help increase the sources of students and assist in obtaining resources and opportunities through cooperations with different entities. Therefore, higher education institutions should actively establish a good brand image, search for positive and negative information that may affect their brand image and control such information to influence customer perceptions of their brand image (Wang, Chen, & Chen, 2012).

Prior research has considered brand images’ links or associations to brands that exist in consumers’ memory, which are then used to infer or maintain the perceived quality of a product and to represent all information regarding that product. Thus, this study follows Park, Jaworski and MacInnis (1986) by categorising the brand images perceived by students-as-customers as functional, symbolic and experiential to evaluate brand image.

The most influential theories regarding brand identification have been proposed by Aaker (1991) and Kepherer (1992). Aaker’s (1996) brand identity system consists of four chief parts: brand identity, value proposition, credibility and brand–customer relationship. Consumers are inclined to select brands that they identify with or that represent them. A branded product that is owned conveys the self-concept of the consumer, and possessing the branded product constitutes a channel for maintaining a positive self-image (Belk, 1988).

Balaji, Roy and Sadeque (2016) identified the role of university brand personality, knowledge and prestige in developing student identification towards a university. In addition, the effects of student–university identification on various university supporting behaviours (i.e. university affiliation, suggestions for improvement, advocacy intentions and participation in future activities) were studied. The research results demonstrated that university brand knowledge and university brand prestige determined student–university identification. In addition, students who identified with their university believed that their future was partially defined by that institution, prompting them to engage in university-supportive behaviours. The results showed that universities should develop brand promotion activities to cultivate students who strongly identify with their university and increase their supporting behaviours towards it.

Consumers who identify with certain brands perceive positive psychological outcomes because of their identity with the brand and thus develop favourable actions regarding the brand. Hence, consumers’ brand identity reveals the intensity of the relationships between consumers and brands (Dimitriadis & Papista, 2010). Kuenzel and Halliday (2010) proposed similar views. To achieve personal growth and satisfy personal esteem, consumers typically like to identify with brands that have positive images. Based on this, if students study in universities with a positive brand image, then those schools can be deemed as advantageous universities, and hence the students will positively identify with these schools. To succeed in the competitive higher education sector, universities should learn from cases of successful promotion. In the business sector, knowledge regarding branding is extensive; however, further research is required in the nonprofit higher education sector. Higher education institutions provide unique services, and a deep understanding of brand identity, meaningfulness, image and prestige within the sector could help brand owners more effectively communicate with relevant stakeholders, including lecturers, students, alumni and employers (Hemsley-Brown, Melewar, Nguyen, & Wilson, 2016). Therefore, this study defines a school’s brand identity as when university students’ self-concept is consistent with that of the university, leading them to formulate a unified cognition with the school.

Kepherer (1992) suggested that, in the identification dimension, brand identification consists of self-image, product attributes, brand personality and the relationship between culture and consumers. A person can possibly become attached to an object because of a trait of that object that is regarded as a self-extension or self-expression (Belk, 1988). Brand identification is distinguished by an intense emotion caused by attachment to the brand that resembles a sense of belonging to the brand (Donavan, Janda, & Suh, 2006). A consumer is attracted by the branded product, and the self-concept, self-image and self-awareness of the consumer can be emotionally connected or associated with the brand through the sentiment established between the brand and the consumer, as well as by the consumer’s perceptions and preferences regarding the image and personality of the brand. In summary, brand identification stems from a connection between the consumer’s self-concept and the brand’s concept or image. The consumer’s sentiment is established based on the perception of the brand image and personality.

Based on the statement above, if students studying at universities have a positive brand image, then the schools are conceived as being advantageous universities; thus, the students will positively identify with these schools. Therefore, this study identifies the operational definition of brand identification as the correlation between university students’ perceptions of school brands and their self-concepts and the degree of sentiment or feeling that is established between schools and students.

Word-of-mouth is the most important informal communication channel for consumers. The reason why WOM is so powerful is that it is a living, direct, experienced and face-to-face process; it has the credibility of information sources and is mainly communicated through friends, relatives, colleagues or well-known experts. Because it is generally easier for people to believe their close ones, WOM affects people’s decision-making process (Wirtz & Chew, 2002).

Rosen (2001) noted that WOM encompasses all reviews about a brand and the sum of all comments about a particular product, service or company communicated among people, and that WOM is different from traditional marketing because it shifts the focus of the market from the relationship between a brand and individual customers to the interaction among customers. Word-of-mouth communication plays a very important role in customers’ consumption behaviour and affects their short- and long-term purchase decisions for a product or service (Assael, 2004).

Word-of-mouth can be divided into input WOM and output WOM in terms of subjects (File, Cermak, & Prince, 1994). Input WOM refers to the receipt of WOM from others before consumption, while output WOM is the WOM of customers after consumption. The WOM communication behaviour of students in this study refers to output WOM, that is, the WOM of university students to other consumers produced via brand image, brand awareness and brand identity after interaction with service staff in the universities.

Word-of-mouth communication entails discussions with others regarding particular products or services in which the messages may be positive or negative (Lin & Chen, 2015). Negative WOM communication refers to consumers communicating their unfavourable purchase experiences to others (Yang & Chou, 2014). Knauer (1992) found that on average one dissatisfied customer is expected to tell nine other customers of his or her own dissatisfied experience; on the contrary, a satisfied customer will tell approximately five other customers of his or her own satisfied experience. Hennig-Thurau, Gwinner and Gremler (2002) also argued that positive WOM communication is an information communication relationship where customers ‘recommend and even show off a relatively excited, lively and novel experience to others’ in the assessment of a product or service. As a result, customers satisfied with the experience of being served will send a positive message to others, leading to positive WOM.

Word-of-mouth communication behaviour of consumers is divided into positive WOM communication and negative WOM communication. This study focuses on the positive WOM communication behaviour of consumers. On the whole, consumers’ positive WOM communication behaviour can be divided into satisfied experience sharing and recommendation to others. When a consumer has brand identity, it means that she or he is very satisfied and will make purchases again, it is easy for him or her to take the initiative to share his or her satisfied consumption experience with others, and she or he will also recommend others to use products or services of that brand.

In summary, WOM communication among consumers can meet the needs of social belonging and identity of consumers. Moreover, it is found through the definition of WOM communication in the literature review and by scholars that WOM communication behaviour can be divided into the two major dimensions of experience sharing and recommendation to others for further discussion. As the content of WOM communication can be positive or negative (Richins, 1983), but only students’ positive WOM communication is really beneficial to schools, this study shall explore positive WOM communication and the two behaviours: satisfied experience sharing and recommendation to others. This study defines WOM communication as university students who share their satisfied learning experience with others and recommend their schools to others.

The hypotheses herein are formulated and extended based on theories selected from the literature review. Subsequent data analyses are conducted to verify these hypotheses. Based on the research objective, the literature review and the research framework, the following hypotheses are set up and empirically verified to describe the correlations between various constructs:

The questionnaire used in this study referenced previous literature and was modified based on the research purposes herein. Each scale underwent a pre-test, as well as reliability and validity tests, and then some items were deleted accordingly. The questionnaire of the pre-test consists of five parts: basic information, brand awareness, brand image, brand identification and WOM communication. The study adopts Keller’s brand awareness scale and takes brand awareness and brand recall as the bases for measurement to learn about the recognition of brand awareness from universities as the research subjects. The scale of the brand awareness followed the questionnaire by Keller (2008), and the eight items were divided into two groups: brand recognition and brand recall. There are nine items about the brand image of higher education institutions, following Park et al. (1986) who divided them into three types: functional image, symbolic image and experience image. After understanding the concept of brand identification by Aaker (1996), the study designed the brand identification scale used herein, and the scale includes six items. The results of the reliability analysis of the variables are as follows: Cronbach’s alpha of brand awareness, brand image, brand identification, sharing of satisfactory experience and recommendation are all greater than 0.7.

This study used a questionnaire survey and brand identification perspective to investigate the influences of brand awareness and brand images on WOM communication for students when selecting and attending universities. The study collected data by distributing online questionnaires through non-random sampling, and the participants were enrolled university students and graduates of public and private universities in Taiwan. In total, 600 questionnaires were sent by e-mail, and 496 questionnaires were returned for a response rate of 82.6%. After encoding, 468 of the collected questionnaires were valid for a valid response rate of 78%.

This study utilised SPSS 21.0 and AMOS 20.0 statistical software packages for the analyses. In addition to the basic multivariate statistical analysis, structural equation modelling (SEM) was performed several times to establish the research model herein and to verify the significance of the direction of influence for the variables and the effectiveness of the research framework. The causal relationships among brand awareness, brand image and WOM communication were also analysed. Furthermore, the study examined whether brand identification mediates the correlation models among brand awareness, brand image and WOM communication.

The participants were enrolled university students and graduates in Taiwan. Among the effective responses, a slim majority of participants were females, totalling 243 participants and occupying 51.93% of the sample. A total of 252 participants studied in private universities, occupying 53.85% of the sample. Furthermore, the majority of the participants had graduated from a university (276 participants occupying 58.98% of the sample), sophomore students in a university totalled 148 participants (31.62% of the sample) and 146 participants were aged 19–23 years (occupying 31.2% of the sample).

This study uses several research approaches. Firstly, we apply exploratory factor analysis (EFA) to derive the underlying dimensions of all items of the research model. Following from EFA, we use confirmatory factor analysis (CFA) and SEM to test the conceptual model, which examines the research structural model (Anderson & Gerbing, 1988; Baumgartner & Homburg, 1996). Implementation of the statistical approach is enabled through the AMOS 20.0 computer program. We utilise CFA to identify the underlying factors from a whole set of items in the study, following the procedure of Steenkamp and Van Trijp (1991) and Gerbing and Hamilton (1996). We use extraction by the principal component method, where the fixed number of factors equals 5, and execute rotation by varimax to identify the latent structure of all items (Hair, Black, Babin, Anderson, & Tatham, 2006). The cut-off factor loading for discarded items is 0.60, while items with cross-loadings above 0.4 are discarded. From the EFA results, we keep 18 items for CFA in the next stage.

This study used CFA to evaluate the factor structure, reliability and validity of the scales. In addition, composite reliability (CR) of the latent variables was used to measure the internal consistency of all variables. A higher CR implies higher internal consistency for the measurement index. All CRs of the latent variables are greater than the minimum threshold of 0.60, showing superior internal quality for the research model.

In Table 1 the individual item reliability demonstrates that the factor loadings of the manifest variables on the latent variables are all greater than 0.650, and that the z-values all achieve a level of significance, thus demonstrating a significant correlation between the manifest variables and their related latent variables. As shown in Table 2, the average variances extracted (AVEs) of the variables are all 0.56 and greater, which is an acceptable level. This indicates strong convergent validity for this study. In addition, in the model the square roots of the AVEs for each latent variable are greater than the correlation coefficients between the variables. Thus, the latent variables herein manifest sufficient discriminant validity.

AMOS 20.0 software was utilised for data analysis to conduct the most typical method of determining SEM goodness-of-fit; that is, the χ² test. In the measurement model, although χ² (46) = 250.446 (p = 0.00 < 0.05), the χ² value is extremely sensitive to a large sample and to a non-normal distribution, and when these conditions occur the χ² value increases. Consequently, based on Bagozzi and Yi (1988), this study adopted residual analysis indices that are less restricted by degrees of freedom (RMSEA = 0.055). In addition, the study used the minimum threshold of 0.08, as advised by Hu and Bentler (1995), as the referential index. Moreover, three indices were used to evaluate the goodness-of-fit of the research model: the goodness-of-fit index (GFI = 0.927) for the absolute fit measures of the overall measurement model, as well as the normed fit index (NFI = 0.957), and the comparative fit index (CFI = 0.978) for incremental fit measures. All three indices surpass 0.90 and are therefore acceptable. The overall model goodness-of-fit is superior because all of the measurement values meet the standard evaluation requirements.

This study formulated SEM using AMOS 20.0 statistical analysis software and conducted two verification processes. Firstly, the study tested the research structural model to precisely evaluate the goodness-of-fit of the overall theoretical model. Secondly, the study verified the influences and results of all latent variables and tested the causal relationships among the latent variables in the structural model to verify the research hypotheses.

The purpose of evaluation regarding overall model goodness-of-fit standards is to see whether a theoretical model explains or fits the observation data. Therefore, this study adopted the evaluation items developed by Bagozzi and Yi (1988) to perform relevant evaluations (Table 3). Although χ² = 309.108 (p = 0.00 < 0.05), the χ² value is extremely sensitive to a large sample and non-normal distribution, and when these conditions occur the χ² value increases. In addition, three indices were used to evaluate the goodness-of-fit of the research model: the goodness-of-fit index (GFI = 0.912) for the absolute fit measures of the overall measurement model, and the normed fit index (NFI = 0.948) and comparative fit index (CFI = 0.967) for the incremental fit measures. All three indices surpass 0.90 and are acceptable. The overall model goodness-of-fit is superior because all of the measurement values meet the standard evaluation requirements.

The method of maximum likelihood estimation (MLE) was used to evaluate this study’s path values and to test whether the study’s hypotheses obtained a level of significance. Table 4 shows the verification results for the research hypotheses.

As shown in Table 4 and Figure 1, the hypothesis path between brand awareness and brand identification is 0.582 (p < 0.001) and significant, indicating that brand awareness significantly and positively influences brand identification, and that brand awareness elevates the relationship proneness of brand identification. Thus, H1 regarding brand awareness significantly and positively influencing brand identification is supported.

The hypothesis path between brand image and brand identification is 0.362 (p < 0.001) and significant, denoting that brand image significantly and positively influences brand identification, and that brand image elevates the relationship proneness of brand identification. Thus, H2 regarding brand image significantly and positively influencing brand identification is supported.

The hypothesis path between brand identification and satisfactory experience sharing is 0.741 (p < 0.001) and significant, implying that brand identification significantly and positively influences satisfactory experience sharing, and that brand identification elevates the relationship proneness of satisfactory experience sharing. Hence, H3 regarding brand identification significantly and positively influencing satisfactory experience sharing is supported.

The hypothesis path between brand identification and recommendations to others is 0.812 (p < 0.001) and also significant, indicating that brand identification significantly and positively influences recommendations to others, and that brand identification elevates the relationship proneness of recommendations to others. Consequently, H4 regarding brand identification significantly and positively influencing recommendations to others is supported.

Most prior studies have referred to the method when verifying mediating effects. However, because of a lack of statistical power, other novel verification methods have been developed (Sobel, 1982), although they are not robust. In recent years, many researchers have proposed the bootstrap method, which has the highest statistical power and robustness for verifying mediating effects. Therefore, this study also adopted the bootstrap method. In the bootstrap method, if zero is not contained within a (1-α) 100% confidence interval (CI; typically set at 95% CI), then statistical significance is achieved at an internally significant standard. The study’s verification results (Table 5) indicate that the mediating effect of brand identification on the influence that brand awareness has on satisfactory experience sharing and recommendations to others achieves significance (95% CI). Similarly, the mediating effect of brand identification on the influence of brand image on satisfactory experience sharing and recommendations to others is significantly different from zero. In summary, H5 (H5a–H5d) is supported.