Antecedent of WOM and Its Implications on The Brand Image of The Higher Education Institution in Bali

The purpose of this study is to explain the Antecedent of WOM and Its Implications on The Brand Image of The Higher Education Institution in Bali. The results showed that the variables tangible, responsiveness, and empathy had a positive and significant effect on satisfaction, while reliability and assurance had a positive and insignificant effect on satisfaction. Furthermore, satisfaction has a positive and significant effect on WOM and on brand image, and WOM has a positive and significant effect on brand image. In the future higher education institutions in Bali should increase reliability and assurance in order to increase satisfaction..


INTRODUCTION
Higher education institutions in Indonesia continue to develop rapidly.This phenomenon is a good thing for prospective students.More and more higher education institutions are providing opportunities for prospective students to determine their choice of choosing the best place to study.The conditions that occur in Indonesia are also similar to those of higher education institutions in Bali.The number of higher education institutions in Bali has also experienced rapid development and this also provides room for prospective students to make decisions about where to study.
The choice of prospective students or the market, of course, is determined by many things.One of the main determining factors is the brand image of the higher education institution which is formed from its experience (Nixon et al., 2018).All higher education institutions strive to enhance their brand image in order to win the competition and become the market's first choice (Nguyen et al., 2019).Brand image is able to provide an overview of the quality of service received (Watjatrakul, 2014;Garipagaoglu, 2016).Currently, the brand image of higher education institutions in Bali is not maximal.Based on data from Kompas.com, the order of universities in Bali is not in the top 10, but at number 42 for Udayana University.This is indicated by the large number of prospective students who wish to study outside Bali.So that Bali becomes worthy of research in order to find out the things that cause prospective students to choose universities outside Bali.Therefore, higher education institutions in Bali really need to pay attention to and maintain their brand image so that it becomes stronger and becomes the first choice for prospective students if they want to study.The brand image of the higher education institution determines its position compared to competitors.(Bennett & Ali-Choudhury, 2009;Syed Alwi & Kitchen, 2014).The brand image of higher education institutions has been widely studied by several previous researchers, such as: Johnson et al. (2001); Casidy (2013); Sultan & Wong (2012, 2013).The brand image reflects what the institution looks like in the eyes of its consumers (Bravo et al., 2012).
Brand image can be formed from positive WOM communication (Reza et al., 2019).The more positive the WOM, the higher the brand image of a higher education institution.Several researchers have examined the relationship between WOM and brand image (Kamboj & Rahman (2017).The results shown are that the better WOM communication from an institution is able to strengthen its brand image (Tran et al., 2015;Yang & Mutum, 2015;and Raji et al. al., 2018).
In addition, to strengthen or improve brand image it is also necessary to pay attention to customer satisfaction (Ngamkamollert & Ruangkanjanases, 2015).If the customer feels satisfied, the brand image of a company will also increase.Research from Ali et al. (2016); Bini & Masserini (2016); Kim & Lee (2015), who conducted research at higher education institutions, found that student satisfaction has a positive and significant effect on brand image, but there are still other researchers such as Andreassen & Lindestad (1998), who show inconsistent results, namely satisfaction.has not been able to build a brand image.In addition, there is also another study that examines the opposite relationship, namely, the effect of brand image on student satisfaction (Brown & Mazzarol, 2009).
Because there is still a reciprocal relationship and there is a research gap on the effect of satisfaction on brand image, it is necessary to supplement it with a mediating variable.The suitable mediating variable to use is the WOM variable.The reasons for entering the WOM variable are motivated by several considerations, including: 1) If only satisfaction is felt by the customer, then the customer keeps that feeling in his heart, It would be nice that customer satisfaction needs to be continued by releasing that feeling through positive communication to others.This is positive WOM that arises from a sense of satisfaction.This can be seen in the results of research by Rubio et al. (2015;2020); 2) besides that, if positive WOM can emerge, it will also be able to improve the brand image of the institution (Kuenzel & Halliday, 2010).
The explanation above illustrates that in order to build a strong brand image of higher education institutions, it is necessary to increase positive WOM and student satisfaction.Student satisfaction is also a very important determining variable in building a brand image.Therefore, higher education institutions must know strategies to increase student satisfaction.A strategy to increase student satisfaction that is commonly implemented is to develop the best service quality (Arokiasamy & Abdullah, 2012;Lupo, 2014;and Štimac & Šimic, 2012).The quality of service is what these higher education institutions can fulfill towards the expectations of their students.Service quality includes: tangible, reliability, assurance, responsiveness, and empathy (Kim et al., 2005).Tangible, such as the availability of ample parking, clean lecture halls, comfortable lecture halls, computer facilities, neat staff appearance, clean toilets, complete book collections, and availability of online learning media, can increase student satisfaction (Abdullah, 2006).Reliability, such as: fast staff service, timely staff service, clear staff service, and careful service, also have a positive and significant effect on student satisfaction.Likewise, responsiveness, assurance, and empathy can increase student satisfaction (Katircioglu et al., 2012).Based on the existing background, this study aims to determine the effect of service quality on student satisfaction, WOM, and the brand image of higher education institutions in Bali.
More and more higher education institutions are providing opportunities for prospective students to determine their choice of choosing the best place to study.The conditions that occur in Indonesia are also similar to those of higher education institutions in Bali.The number of higher education institutions in Bali has also experienced rapid development and this also provides room for prospective students to make decisions about where to study.
The choice of prospective students or the market, of course, is determined by many things.One of the main determining factors is the brand image of the higher education institution which is formed from its experience (Nixon et al., 2018).All higher education institutions strive to enhance their brand image in order to win the competition and become the market's first choice (Nguyen et al., 2019).Brand image is able to provide an overview of the quality of service received (Watjatrakul, 2014;Garipagaoglu, 2016).Currently, the brand image of higher education institutions in Bali is not maximal.This is indicated by the large number of prospective students who wish to study outside Bali.Therefore, higher education institutions in Bali really need to pay attention to and maintain their brand image so that it becomes stronger and becomes the first choice for prospective students if they want to study.The brand image of the higher education institution determines its position compared to competitors.(Bennett & Ali-Choudhury, 2009;Syed Alwi & Kitchen, 2014).The brand image of higher education institutions has been widely studied by several previous researchers, such as: Johnson et al. (2001); Casidy (2013); Sultan & Wong (2012, 2013).The brand image reflects what the institution looks like in the eyes of its consumers (Bravo et al., 2012).
Brand image can be formed from positive WOM communication (Reza et al., 2019).The more positive the WOM, the higher the brand image of a higher education institution.Several researchers have examined the relationship between WOM and brand image (Kamboj & Rahman (2017).The results shown are that the better WOM communication from an institution is able to strengthen its brand image (Tran et al., 2015;Yang & Mutum, 2015;and Raji et al. al., 2018).
In addition, to strengthen or improve brand image it is also necessary to pay attention to customer satisfaction (Ngamkamollert & Ruangkanjanases, 2015).If the customer feels satisfied, the brand image of a company will also increase.Research from Ali et al. (2016); Bini & Masserini (2016); Kim & Lee (2015), who conducted research at higher education institutions, found that student satisfaction has a positive and significant effect on brand image, but there are still other researchers such as Andreassen & Lindestad (1998), who show inconsistent results, namely satisfaction.has not been able to build a brand image.In addition, there is also another study that examines the opposite relationship, namely, the effect of brand image on student satisfaction (Brown & Mazzarol, 2009).
Because there is still a reciprocal relationship and there is a research gap on the effect of satisfaction on brand image, it is necessary to supplement it with a mediating variable.The suitable mediating variable to use is the WOM variable.The reasons for entering the WOM variable are motivated by several considerations, including: 1) If only satisfaction is felt by the customer, then the customer keeps that feeling in his heart, It would be nice that customer satisfaction needs to be continued by releasing that feeling through positive communication to others.This is positive WOM that arises from a sense of satisfaction.This can be seen in the results of research by Rubio et al. (2015;2020); 2) besides that, if positive WOM can emerge, it will also be able to improve the brand image of the institution (Kuenzel & Halliday, 2010).
The explanation above illustrates that in order to build a strong brand image of higher education institutions, it is necessary to increase positive WOM and student satisfaction.Student satisfaction is also a very important determining variable in building a brand image.Therefore, higher education institutions must know strategies to increase student satisfaction.A strategy to increase student satisfaction that is commonly implemented is to develop the best service quality (Arokiasamy & Abdullah, 2012;Lupo, 2014;and Štimac & Šimic, 2012).The quality of service is what these higher education institutions can fulfill towards the expectations of their students.Service quality includes: tangible, reliability, assurance, responsiveness, and empathy (Kim et al., 2005).Tangible, such as the availability of ample parking, clean lecture halls, comfortable lecture halls, computer facilities, neat staff appearance, clean toilets, complete book collections, and availability of online learning media, can increase student satisfaction (Abdullah, 2006).Reliability, such as: fast staff service, timely staff service, clear staff service, and careful service, also have a positive and significant effect on student satisfaction.Likewise, responsiveness, assurance, and empathy can increase student satisfaction (Katircioglu et al., 2012).Based on the existing background, this study aims to determine the effect of service quality on student satisfaction, WOM, and the brand image of higher education institutions in Bali.The novelty of this study is, this research provides an update in terms of the research location, namely the choice of Bali as the research location in researching the brand image of universities in Bali, because Bali was previously more famous for tourism, not education.

Hypothesis Development Tangible Influence on Student Satisfaction
According to the research results of Ryu & Han (2010), it shows that tangible variables have a positive and significant effect on student satisfaction.Furthermore, Ryu et al. (2012), also examined the effect of tangibles on student satisfaction, which showed that better physical evidence was able to increase student satisfaction.Tangible shown by: a large parking area, comfortable lecture rooms, clean toilets, a complete collection of books, the appearance of employees who are getting better is able to make students feel satisfied.This is also supported by several other researchers, such as Han & Ryu (2009); Ali et al. (2016); and Cham et al. (2016).Based on the results of empirical studies, the following hypothesis can be built: H1: Tangible has a positive and significant effect on student satisfaction The effect of Reliability on satisfaction Reliability possessed by education personnel in speed, accuracy, and clarity in serving students can increase student satisfaction.The effect of reliability on student satisfaction has been studied by Zeithaml et al. (2008).The results of his research show that the higher the level of reliability, the higher the student's satisfaction.The same thing is also shown by the research results of Richardson (2005) and Katircioglu et al. (2012).Based on the results of empirical studies, the following hypotheses can be formulated: H2: Reliability has a positive and significant effect on student satisfaction The effect of responsiveness on student satisfaction Responsiveness which is shown by the responsiveness of the education staff is able to increase student satisfaction.The effect of responsiveness on satisfaction has been studied by several researchers, including: Abdullah (2006); Arokiasamy & Abdullah (2012); Katircioglu et al. (2012); Ali et al. (2016); The results of his research show that the higher the responsiveness, the higher the student's satisfaction.Based on the results of empirical studies, the following hypotheses can be formulated: H3: Responsiveness has a positive and significant effect on student satisfaction The effect of assurance on student satisfaction Assurance can also increase student satisfaction (Sultan & Wong, 2012).The results showed that assurance such as time assurance regarding the administrative process, guarantee of the best service can increase student satisfaction.Furthermore, the results of research by Sultan & Wong (2013) also show consistent results, namely assurance can increase student satisfaction.Sharifah et al. ( 2019), examining service quality, which includes the assurance variable, shows the results that assurance can lead to higher student satisfaction.Based on the results of empirical studies, the following hypothesis can be built: H4: Assurance has a positive and significant effect on student satisfaction.

The effect of empathy on student satisfaction
Empathy is usually indicated by an indication of special attention to customers or paying attention to students who have physical or nonphysical weaknesses which can also increase student satisfaction (Arokiasamy & Abdullah, 2012).This was revealed in the results of research conducted by Bini & Masserini (2016), which stated that by giving empathy, higher attention from the heart to students would increase their satisfaction.Likewise, the research results of Voss et al. (2010) examining the influence of the empathy variable on student satisfaction stated that the results of their study showed empathy had a positive and significant effect on student satisfaction.Based on the results of empirical studies, the following hypothesis can be built: H5: Empathy has a positive and significant effect on student satisfaction The effect of student satisfaction on WOM Student satisfaction can also increase positive WOM (Rojas-Mendez et al., 2009).This is revealed in the results of research conducted by Paswan & Ganesh (2009), which state that student satisfaction can increase positive WOM.Furthermore, the research results of Harrison & Shaw (2004); Rubio et al. (2015;2020), also showed consistent results, namely student satisfaction was able to increase positive WOM.Based on the results of existing empirical studies, the following hypothesis can be built: H6: Student satisfaction has a positive and significant effect on WOM The effect of student satisfaction on brand image Student satisfaction can also improve brand image (Palmer et al., 2016;Arambewela & Hall, 2009).This is revealed in the results of research conducted by Kuenzel & Halliday (2010), which states that student satisfaction improves the brand image of higher education institutions.Furthermore, the research results of Ali et al. (2016); Bini & Masserini (2016);and Sharifah et al. (2019), also showed consistent results, namely student satisfaction was able to improve brand image.Based on the results of empirical studies, the following hypothesis can be built: H7: Student satisfaction has a positive and significant effect on brand image The effect of WOM on brand image WOM can also improve brand image (Kuenzel & Halliday, 2010), which states that positive WOM can strengthen brand image.Likewise, the results of research that are consistent with examining the effect of WOM variables on brand image, such as: Yang & Mutum (2015); and Jalilvand & Samiei (2020), reinforce the results of research which show that WOM or e-WOM is able to improve the brand image of a product.Based on the results of empirical studies, the following hypothesis can be built: H8: WOM has a positive and significant effect on brand image The role of WOM in mediating student satisfaction variable with brand image Research conducted by Garipagaoglu (2016) states that the image of higher education institutions is very important to note.Therefore, a brand image should be built with various efforts, including by providing satisfaction to students and building a positive WOM.Kim and Lee (2015) state that student satisfaction can improve the brand image of higher education institutions.Likewise, the increasingly positive WOM is able to improve the brand image of higher education institutions (Yang & Mutum (2015).So the position of WOM can be used as a mediating variable between student satisfaction and brand image.Based on the results of empirical studies, the following hypothesis can be formulated: H9: WOM is able to significantly mediate student satisfaction with brand image

METHOD
This research is classified into associative research, namely research that aims to determine the relationship between tangible variables, reliability, assurance, responsiveness, and empathy, student satisfaction, WOM, and the brand image of higher education institutions in Bali.Data was collected through an online questionnaire (Google Form), the respondents of this study were all prospective students in Bali.The number of respondent is 370 respondent.
In this study, there are three types of variable data to be studied, these variables are: Exogenous variables: Tangible, Reliability, Responsiveness, Assurance, Empathy; Mediating variables: Satisfaction; WOM; and Endogenous Variable: Brand Image.
The variable indicators of the tangible variables, reliability, assurance, responsiveness, empathy, satisfaction, WOM, brand image were measured by the perceptions of active students from higher education institutions in Bali using a five-level Likert scale, namely from strongly disagree = 1, disagree = 2, quite agree = 3, agree = 4, and strongly agree = 5.
The instrument used was tested for validity and reliability in order to measure what you wanted to measure and to know the consistency of the responses given by the respondent.Testing the validity of the instrument using Pearson's Product Moment correlation technique with a minimum limit of r = 0.3.Instrument reliability testing is done by calculating the reliability coefficient of Cronbach's Alpha with the minimum alpha coefficient> 0.7.Both tests use the SPSS computer program.This analysis is used to describe the research variables, without drawing generalizations.The data that has been collected is then tabulated in a table and a descriptive discussion is carried out.The data analysis technique used in this research is path analysis using SEM -PLS.The use of two different

RESULT AND DISCUSSION
After doing the research, the results can be presented as follows.The characteristics of respondents according to demographic characteristics are presented in Table 2.
Table 2. explains that there were more female respondents than male respondents where female respondents were 62.7 percent while male respondents were 37.3 percent.When viewed by age, the majority of respondents aged 18-26 were 86.49percent.Based on the field of student science, it can be seen that as many as 323 people with a percentage of 87.30 percent are students studying in the socio-cultural field and 47 people with a percentage of 12.7 percent are students in the field of technology science.Furthermore, seen from the status of the tertiary institutions where they study, there were 272 people with a percentage of 73.5 percent from private universities and 98 people with a percentage of 26.5 percent from state university students.
Testing the validity and reliability of the instrument was carried out with Pearson's correlation and Cronbach's Alpha coefficient.The results of testing the validity and reliability of the instrument for each variable can be presented in Table 3.The results of testing the validity and reliability of the instrument in Table 3 show that all variables are valid because the correlation value is above 0.30 and reliable because the Cronbach's Alpha value is above 0.6.Furthermore, the results of variable descriptive analysis explain the respondents' perceptions of tangible variables, reliability, assurance, responsiveness, empathy, satisfaction, WOM, and brand image obtained through distributing questionnaires using google form.Quantitative assessment uses an interval scale by integrating the average score according to the rating category.

PLS SEM Analysis Results
This study uses a two-stage approach to measuring the model before it is used for hypothesis testing, which aims to verify the validity and reliability of a research model.First, by ana-lysing convergent validity, followed by analysing discriminant validity.

Outer Model Test Convergent Validity
The outer model test is carried out to ensure the research indicators are feasible to use as their role in measuring the research variables, so to see whether a model is valid to be the basis for research, there are three criteria that must be met, namely: (1) all loading indicators must be above 0.65; (2) composite reliability (CR) must be above 0.8, and (3) average variance extracted (AVE) for each construct must exceed 0.5..828

Reliability
Based on Table 4, it shows that all outer loading indicators have a value above 0.65 with a range between 0.671 to 0.935 meaning that they are at the recommended limit, then the composite reliability (CR) value is in the range between 0.912 to 0.958, all of which are above 0.8 meaning that all constructs are formed to have good consistency as a research model, the third is the average variance extracted (AVE) value where all values are above 0.5, namely with a range from 0.583 to 0.849 so it can be concluded that the research model in this study has good validity.

Discriminant Validity
To evaluate discriminant validity, a research model is recommended to ensure that the root average variance extracted (√AVE) value of a latent variable must be greater.Discriminant validity is considered good if the smallest AVE root value (√AVE) in Table 6 is greater than the correlation value between the highest latent variables in Table 8.The research model proposed in this study can be considered good, where the smallest √ AVE value is 0.763 more greater than the correlation value between the highest latent variables which is equal to 0.710.

Inner Model Test
Structural models focus on the hypothesized relationships or pathways between latent variables.The results of the inner model test can be seen in Figure 2.

Hypothesis testing
The significance of the estimated parameters provides very useful information about the relationship between the research variables.The basis used in testing the hypothesis is the value contained in the output path coefficients which are presented in Table 7.
Hypothesis testing is done using t-statistics and looking at the p-value.If the p-value <0.05, the hypothesis is accepted.Based on Table 7, it can be explained that tangible satisfaction has a t-statistic value of 2.273 with a p-value of 0.023 <0.05, so the hypothesis is accepted.This means that the better the tangible conditions, the higher student satisfaction.Reliability to student satisfaction has a t-statistic value of 0.254 with a p-value of 0.800> 0.05, so the hypothesis is rejected.This means that the higher the level of reliability, it is not able to increase student satisfaction.Responsiveness to student satisfaction has a tstatistic value of 0.888 with a p value of 0.375> 0.05, so the hypothesis is rejected.This means, the higher the level of responsiveness is not able to increase student satisfaction.Assurance on student satisfaction has a t-statistic value of 5,323 with a p value of 0.000 <0.05, so the hypothesis is accepted.This means, the higher the level of assurance, the student satisfaction will increase.
Empathy towards student satisfaction has a t-statistic value of 6,129 with a p value of 0.000 <0.05, so the hypothesis is accepted.This means that the higher the empathy level, the higher student satisfaction.Student satisfaction with WOM has a t-statistic value of 33,673 with a p value of 0.000 <0.05, so the hypothesis is accepted.This means that the higher the level of student satisfaction, the higher the positive WOM.Student satisfaction with brand image has a t-statistic value of 5,163 with a p value of 0.000 <0.05, so the hypothesis is accepted.This means that the higher the level of student satisfaction, the higher the brand image.Furthermore, WOM on brand image has a t-statistic value of 7,759 with a p value of 0.000 <0.05, so the hypothesis is accepted.This means that the higher the positive WOM, the higher the brand image.

Indirect Effect Testing (Mediation Test)
Test the role of WOM mediation on the effect of satisfaction on brand image by examining the indirect effects which are the Smart PLS output as presented in Table 8.Based on Table 8, it can be explained that the t-statistic value is greater than the t-table value (7,750> 1.96), then WOM significantly mediates satisfaction with brand image.

Discussions
Based on the results of the analysis of the tangible effect on student satisfaction, the beta coefficient value is 0.135 with a significance level of 0.023 ≤ 0.05, which means that H1 is accepted.These results mean that the tangible variable has a positive and significant effect on student satisfaction.So, the better the tangible that educational institutions in Bali have, which is shown by the existence of clean lecture rooms, comfortable lecture rooms, large parking lots, clean toilets, neat appearance of educational staff, having online learning media, the student satisfaction will increase.The results of this study at the same time strengthen the results of previous studies conducted by Ali et al. (2016);and Cham et al. (2016), which states that tangible has a positive and significant effect on student satisfaction.Furthermore, the results of the analysis of the effect of reliability on student satisfaction obtained a beta coefficient of 0.013 with a significance level of 0.800> 0.05, which means that H2 is rejected.This result means that the variable reliability has a positive and insignificant effect on student satisfaction.So, the better the reliability of educational institutions in Bali, which is shown by the service of fast educational staff, fast academic services, clear services, and careful academic services, then it has not been able to increase student satisfaction.Based on the results of the analysis of the effect of assurance on student satisfaction, the beta coefficient value is 0.368 with a significance level of 0.000 ≤ 0.05, which means that H4 is accepted.This result means that the assurance variable has a positive and significant effect on student satisfaction.So, the better the assurance that is owned by educational institutions in Bali which is shown by skill in serving, accuracy in serving, guarantee of friendliness of service, assurance of lecturer competence, and assurance of guiding well, the student satisfaction will increase.The results of the analysis of the effect of responsiveness on student satisfaction obtained a beta coefficient value of 0.047 with a significance level of 0.375> 0.05, which means that H3 is rejected.These results mean, the res-ponsiveness variable has a positive and insignificant effect on student satisfaction.So, the better the responsiveness of educational institutions in Bali, which is shown to be responsive to student complaints, the less likely it is to increase student satisfaction.Based on the results of the analysis of the effect of empathy on student satisfaction, the beta coefficient value is 0.373 with a significance level of 0.000 ≤ 0.05, which means that H5 is accepted.These results mean, the empathy variable has a positive and significant effect on student satisfaction.So, the better the empathy possessed by educational institutions in Bali which is shown special attention to students, the more student satisfaction increases.
The results of this study at the same time strengthen the results of previous research conducted by Bini & Masserini (2016), which states that empathy has a positive and significant effect on student satisfaction.Furthermore, the results of the analysis of the effect of satisfaction on WOM obtained a beta coefficient value of 0.827 with a significance level of 0.000 ≤ 0.05, which means that H6is accepted.These results mean, the student satisfaction variable has a positive and significant effect on WOM.So, the higher the student's satisfaction which is indicated by the feeling of being happy to attend lectures and being satisfied with all the services available, the positive WOM will increase.The results of this study at the same time strengthen the results of previous studies conducted by Rubio et al. (2015), which states that student satisfaction has a positive and significant effect on WOM, this means that with increasing student satisfaction, it will increase positive WOM about higher education institutions in Bali.The results of the analysis of the effect of satisfaction on brand image obtained a beta coefficient value of 0.370 with a significance level of 0.000 ≤ 0.05, which means that H7 is accepted.
These results mean, the student satisfaction variable has a positive and significant effect on brand image.So, the higher the student's satisfaction which is indicated by the feeling of being happy to attend lectures and being satisfied with all the services available, the brand image will increase.The results of this study at the same time strengthen the results of previous studies conducted by Ali et al. (2016); Bini & Masserini (2016), which states that student satisfaction has a positive and significant effect on brand image.
This result means that the increasing student satisfaction will improve the brand image of higher education institutions in Bali.Furthermore, the results of the analysis of the influence of WOM on brand image obtained a beta coefficient value of 0.522 with a significance level of 0.000 ≤ 0.05, which means that H8 is accepted.These results mean, the WOM variable has a positive and significant effect on brand image.So, the better WOM is shown by positive communication about the study program, recommending study programs to others, and the study program is the right choice for study, the brand image will increase.The results of this study at the same time strengthen the results of previous research conducted by Yang and Mutum (2015), which states that WOM has a positive and significant effect on brand image, this means that, with higher positive WOM, it will improve the brand image of higher education institutions in Bali..Based on the analysis of the role of WOM mediation on the effect of student satisfaction on brand image, the beta coefficient value is 0.432 with a significance level of 0.000 ≤ 0.05, which means that H9 is accepted.These results mean, the WOM variable is able to partially mediate the effect of student satisfaction on brand image.So, the effect of student satisfaction as indicated by the feeling of being happy to attend lectures, being satisfied with all the services received on the university's brand image is getting stronger with the positive WOM.

Limitations
As stated in the analysis and discussion, this study can theoretically find a relationship between latent variables of service quality on student satisfaction and WOM and brand image.This research can contribute thoughts to leaders of higher education institutions in Bali that building the best service quality is the right strategy to increase student satisfaction, positive WOM, and a strong brand image.
Various limitations were also found in this study mainly due to the following factors.1) This study only uses respondents from a student perspective.2).The mediating variable studied is the WOM variable which turns out that in this study the mediation variable is partially so that in the future it is necessary to think about adding the mediating variables such as student commitment and student confidence.3) Apart from that, it can also be studied from a lecturer perspective.

CONCLUSION AND RECOMMENDATION
Based on the research results, it can be concluded that two factors of service quality that are not able to increase student satisfaction, namely reliability and responsiveness, while tangible, assurance, and empathy are able to increase student satisfaction.Furthermore, student satisfaction can increase positive WOM satisfaction and brand image of higher education institutions in Bali.

Table 2 .
Characteristics of Respondents

Table 3 .
Results of Testing the Validity and Reliability of the Instrument

Table 4 .
Model Size Results

Table 5 .
Correlation Between Latent Variables

Table 6 .
AVE Root Value

Table 8 .
Indirect Effects Correlation Path coefficient t-Statistic p-Values