Analysis of the Relationship Between Image, Social Media, and Attitude to Predict Intention to Choose: An Empirical Investigation of Presidential Election in Indonesia

Research focusing on presidential, regional, and legislative elections has been conducted by previous researchers. However, there is little research comparing business and non-business respondents in predicting presidential intentions. Therefore, this study aims to compare student respondents from business schools and non-business students in predicting the intention of choosing a president. Moreover, Indonesia will undergo presidential election in 2019 where young voters such as college students are voters in significant amounts. Young voters are often identified as swing voters which their attitudes toward the candidates they choose are influenced by information from social media and the image of the presidential candidate. This study involved 214 respondents from both science and non-science education background. The data was analyzed by applying the structural equation modeling. The results of the study showed that there was no difference in study results for respondents with scientific backgrounds and respondents with non-scientific backgrounds. Both types of respondents show the same results that only 1 dimension of the variable of image that has a positive relationship with the attitude towards the candidate candidates. This dimension is charm. The result also shows that there is significant relationship between attitudes toward choosing the candidate with intention to choose the candidate. This study provides results discussion, managerial implications, and recommendation for future research.


INTRODUCTION
The year of 2018-2019 is a political year that "worries" many political parties in Indonesia. This is because in 2018, Indonesia held 171 direct elections to the regions. Continue to the year 2019 which related to the election of legislative members and the main focus in the year of 2019 is the presidential election. Presidential election is important because this election is closely tied to the future of the Indonesian nation. The wrong choice of the highest leaders produces an unclear future for the people of Indonesia. The importance of the 2019 presidential election is also due to the polarization in civil society that has taken place since the 2014 Presidential Election and was sharpened in the election of the governor in 2017 (Syahputra, 2017;Mursid, 2108).
Young voters (i.e., 17-38 years) are one of the main components in the presidential election, especially in Indonesia. In particular, young voters reach 55% of all presidential voters in 2019 (Ramadhan, 2017). Young voters often become a major concern for parties, legislators, or for presidential candidates themselves. The main characteristics of young voters are that they are critical but unstable and easy to move choices. They are identified as swing voters (Kertati, 2018). Many ways are done to get votes from young people. For example, Prabowo Subianto (Indonesian presidential candidate in 2019) chose Sandiaga Uno as a young vice presidential candidate who is expected to meet millennial expectations. Sandiaga Uno is the deputy governor of the city of Jakarta who later resigned to focus on this presidential election.
Likewise, Joko Widodo (hereinafter referred to as Jokowi), who is also a candidate for Indonesian presidential candidate in 2019, chooses Erick Thohir as the head of his campaign team. Erick Thohir, as a young generation, who is a businessman and founder of the Mahaka Group which is the parent company of a company that has a focus on media and entertainment businesses. He is also the owner of the Inter Milan club based in Italy.
Research in the context of presidential, legislative or governor elections has been conducted in previous studies (e.g., Sulistyo and Poespawardaja, 2016;Warner and Banwart, 2016). In the context of elections, social media is one of the variables that are often used to predict intentions or behavior of voting for young voters (e.g., Perangin-Angin and Zainal, 2018;Syahputra, 2017;Sulistyo and Poespawardaja, 2016;Atmodjo, 2014). However, not many studies have applied candidate image variables as also one of the main variables in predicting voting intentions. In fact, the image of a good candidate will shape the positive attitude of voters who then tend to choose the candidate. Furthermore, relating with young voters, as far as researchers understand that there is no research that tries to compare student respondents as young voters with business and non-business faculty backgrounds as business and non-business students such as engineering students have their own distinctive characteristics. Several previous studies in several contexts indicate that there are significant differences between science students and non-science students (e.g., Asniar, 2016;Takeuchi et al., 2015;Sheeja, 2010;Anaam, 1997;Gogolin and Swartz, 1992). Based on the explanation above, research on youth intentions to choose presidential candidates is important and necessary, thus this research can contribute to theory and practice.

Behavioral Intention and Attitude
Intention is one of the main predictors of behavior. In other words, people's behavior such as buying goods, carrying out activities, and other behaviors can be predicted through their intentions. Many studies in the social field use intention as one of the main dependent variables. In attitude theory developed by Fishbein and Ajzen (1975) explains that a person's behavior is influenced by his intention to behave. The intention is influenced by 3 main variables, namely: attitude, subjective norms, and perceived behavioral control. This study uses attitude as one of the main variables in predicting behavioral intentions where many previous studies support a significant relationship between attitudes and intentions (e.g., Sihombing, 2017;Nagarathanam and Buang, 2016;Suki and Suki, 2011;Sihombing, 2009). Furthermore, previous research supports the stable relationship between attitude and intention (e.g., Chatzisarantis et al., 2005;Rosenthal, 1993;Doll and Ajzen, 1992).
Attitudes can be defined as an evaluation of a person towards a thing such as people, activities, goods, and many others (Ajzen, 1988). This evaluation is formed through many sources such as personal experience, the Internet, friends, family, and many others. Because attitude is an evaluation that comes from the learning process (from a variety of media and experience) so that one of the main characters of the attitude is stable, not easy to change. In relating with elections, understanding candidate evaluations is important because they have a direct and positive effect on vote choice in the elections. It can be stated that the more positively voter evaluates the candidate; the more likely she is to vote for the candidate's. The importance of attitude reflected in any research in the fields of marketing, consumer behavior, and other fields in the social context that uses attitude as one of the main variables in explaining many behavioral contexts (Hooghe and Wilkenfeld, 2008;Bagozzi, 1992).

Candidate Image
Candidate image refers to clusters of voter perceptions of candidates (Hacker, 2004 cited by Warner andBanwart, 2016). Furthermore, this image is more oriented towards the candidate's personality. Furthermore, the candidate image is a cognitive representation in the voters' perception process of the candidate's message (Warner and Banwart, 2016). Warner and Banwart (2016) also show that voters are easier to draw conclusions about the image than to hear the presidential policy debates. Thus, it can be said that the image of the candidate is one of the main factors in shaping the attitude of voters against the presidential candidate.
This research applied a multifactor approach to candidate image based on Warner and Banwart (2016). Specifically, candidate image includes 6 facets of image: character, intelligence, leadership, benevolence, homophily, and charm. Characters refer to quality within people that enables people to live by their values (Opatha, 2007). Another facet of image, namely intelligence, is related to the ability to process information in a way that allows individuals to adapt to their environment (Cohen, 2016). Leadership is conceptually defined as a process of social influence in which a person can ask for help and support from others in achieving common tasks (Chemmers, 2014). Benevolence is another facet of image. The literally means of benevolence is "to wish one well" (Suhr, 2011). The benevolence concept also refers to a genuine desire to help others (Warner and Banwart, 2016). Homophily is defined as an understanding of the people's values and concerns (Prysby, 2007 cited by Warner and Banwart, 2016). The final facet of image is charm. Based on the Oxford dictionariy, charm is defined as the power or quality of delighting, attracting, or fascinating others. Charm can be understood as the likability of a person (Warner and Banwart, 2016). Charm is important in the facets of image because politics in many countries are now changing from focus on parties to individuals or his/her appearances and personalities (Hoegg and Kewis, 2011).

Information from Social Media
Social media terminology consists of two words, namely "media" and "social." Media is a means of communication and the word of social refers to conditions related to society. Thus, social media is a communication tool that is closely related to society. In particular, social media makes people interact through the use of technology and web-based. Furthermore, the use of technology leads to more interactive communication (Baruah, 2012;Kaplan and Haenlein, 2010).
Regarding politics, social media is a main tool used to mobilize mass support. Through various types of social media such as social networking sites, forums or online messages, content communities, and various other types, political communication is delivered by each of the interested parties in this election. Young voters who are millennials are generations that cannot be separated from technology, especially from their smart phones. Through social media, young voters have a lot of information about the candidates they will choose in the presidential election later (Bolton et al., 2013;Chan-Olmstead et al., 2013).

Proposed Research Model
With reference to the theoretical background and the results of previous studies, the following models and hypotheses in this study. Figure 1 describes this research model. Based on the model, the hypotheses in this study are: H 1a : There is a positive relationship between candidate character and attitude towards choosing a candidate. H 1b : There is a positive relationship between candidate intelligence and attitude towards choosing a candidate. H 1c : There is a positive relationship between candidate leadership and attitude towards choosing a candidate. H 1d : There is a positive relationship between candidate benevolence and attitude towards choosing a candidate. H 1e : There is a positive relationship between candidate homophily and attitude towards choosing a candidate. H 1f : There is a positive relationship between candidate charm and attitude towards choosing a candidate. H 2 : There is a positive relationship between information from social media and attitudes toward choosing a candidate. H 3 : There is a positive relationship between attitudes towards choosing candidates and the intention to choose a candidate.
H 4 : There are differences in research results between science students and non-science students.

Sampling Design
This research is a descriptive study where one of the criteria of descriptive research is to identify meaningful relationships between variables in explaining the phenomenon (Hair et al., 2009). Data were collected from questionnaires distributed by nonprobability sampling. Specifically, the sampling type used is the type of judgment sampling, which where the research respondents are individuals who meet the research criteria. These criteria are respondents who are students and come from non-science faculties (for non-science respondents) and come from the science faculty (for science respondents). All respondents are young people who have suffrage in 2019 at the time of the presidential election.

Measurement of Variables
Indicator variables used in this study follow previous studies which have also been validated in previous studies. In particular, indicators for character, intelligence, leadership, benevolence, homophily, and charm variables come from Warner and Banwart (2016). Furthermore, indicators for information variables from social media come from Hsu et al. (2015), and the attitude and intention variable to elect the president is based on Ajzen (2006 "5 = strongly agree." The goodness of measurement was examined by applying reliability and validity analysis. In this study, reliability testing using the composite reliability and average variance extracted (AVE). Specifically, the threshold of 0.7 for composite reliability (Fornell and Bookstein, 1982) and 0.5 for AVE was applied in this study (Fornell and Larcker, 1981).
The validity test used in this study is construct validity which consists of two types of validities namely convergent validity and discriminant validity (Sekaran and Bougie, 2016). Based on Fornell and Larcker (1981), convergent validity was examined through the factor loadings or the regression weight, composite reliability, and AVE. Hair et al. (2006) recommend the threshold of 0.5 for the factor loadings. Another element of construct validity is discriminant validity. Discriminant validity was achieved when the correlation between two factors is lower than 0.85 (Kenny, 2016). Discriminant validity was also achieved when the square root of the AVE is greater than the correlation between constructs (Fornell and Larcker, 1981). Furthermore, Fornell and Larcker (1981) also pointed out that convergent validity is evidenced by a factor shows convergent validity if AVE ≥ 0.50.

Data Analysis
This research applied structural equation modeling (SEM) with maximum likelihood estimation to test research hypotheses. This method was used because SEM has ability to assess the relationships comprehensively (Hair et al., 2006). As suggested by Anderson and Gerbing (1988), the structural equation analysis in this research was conducted in two main stages: the estimation of the measuring model and the estimation of the structural model. In the particular, the former was conducted to confirm the suitability of the proposed scales using the criteria of reliability and validity. Then, the latter was conducted to test the relationships between the constructs.

RESULTS
A total of 214 questionnaires were returned out of 250 distributed, yielding a response rate of 85.6%. As shown in Table 1, the profile of the total sample reveals that female constituted about 60.7% of the sample. However, for non-science students, male represent almost 60% of the sample. The results of the study show that the respondents' preference for the 2019 presidential candidates to come is Jokowi (90%), the current incumbent president.
Confirmatory factor analysis was performed to assess the measurement model for both science (Table 2a) and non-science respondents (Table 2b). For science respondents, the fit indices show a good-fit model (CMIN/DF = 1.268, CFI = 0.966, TLI = 0.955, RMSEA = 0.055). Furthermore, for nonscience respondents, the fit indices also show a good-fit model (CMIN/DF = 1.195, CFI = 0.981, TLI = 0.974, RMSEA = 0.040). Table 2a and b also show that convergent validity was achieved because all measures exceed the recommended threshold which factor loadings ranges from 0.619 to 939, the composite reliability ranges between 0.748 and 0.936, and the AVE ranges between 0.501 and 0.816.
Discriminant validity was examined to assess the construct is distinct from other constructs. Table 3a and b show that discriminant validity is achieved because the correlation between two constructs is lower than 0.85. Next, the square root of the average variance is extracted greater than the correlation between constructs.
After assessing the reliability and validity of all measures, the next step is to examine a hypothesis test. SEM is applied to test the structural relationships between variables. The results of SEM analysis are presented in Table 4a

DISCUSSION AND CONCLUSION
The main objective of this research is to test the structural relationship between candidate image, information from social media, attitude, and intention to choose the president candidate.
The results show that three out of eight hypotheses were supported for both science and non-science respondents.
In conjunction with the hypothesis that states the relationship between prospective image factors with the attitude of choosing a presidential candidate, the results of this study indicate that only 1 factor of image (ie, charm) influences the attitude of choosing a presidential candidate. These results are similar both from respondents with a background in science and not science. Charm is defined as one's attraction. A person's charm can come from various factors. For example, because of his character, intelligence, style of leadership, benevolence and other factors.
In relating with this study, more than 85% of respondents chose Jokowi as a candidate for presidential election in 2019. Jokowi's charm can be explained from several factors. First, his charm comes from at his leadership style which is populist, informal, and even impressed like a villager. However, the community is fascinated by Jokowi's style that is able to communicate with all walks of life without looking fake and without borders. The community will not suspect Jokowi is doing imaging in his daily activities. This is because in the public perception, Jokowi's daily life is like that. This charm makes Jokowi look authentic in the public eye.
Second, Jokowi's charm is also influenced his intelligence. In relation to politics, many parties praised Jokowi's intelligence in responding to criticism of his political opponents. His ability to respond is the result of his intelligence in making opponent's  radicalism, extremism, intolerance, sectarianism and racism in Indonesia, which is a country with a Muslim majority population (Jegho, 2018).
Jokowi's charm is inseparable from the perception of the people who consider himself a popular president. Television or print media in Indonesia often report how Jokowi can mingle with his people on various occasions: at the Palace, on the market, on work visits, and many other places. Not only that, the public acknowledged how no siren sounded when Jokowi was on the highway. Jokowi and his bodyguards can pass on the highway silently unlike previous presidents in power in Indonesia. In short, there is no distance between the president and the people that makes Jokowi's  into an open problem. Specifically, Jokowi often reverses the accusations of his political opponents with soft and humor. This is acknowledged by many observers as an intelligent response (Muhammad, 2017).
Still in relation to politics, many things can show Jokowi's intelligence in politics. Jokowi in 2018 surprised many people in Indonesia, when he as the next presidential candidate in 2019 declared a prominent Muslim cleric as his partner in the upcoming 2019 presidential election. Many people are fully aware that with Ma'ruf as vice president, Jokowi will find it easier to handle many problems related to religion such as  charm able to make a significant difference compared to the charms of previous presidents in Indonesia.
The insignificance of the relationship between image factors such as intelligence, character, leadership, virtue, and homophile can be explained that all of these factors seem to merge into a factor of charm. In other words, Jokowi's charm reflects his intelligence, character, leadership style, and so on as well as previously explained.
In relation to the relationship between attitudes and intention to choose candidates, the two types of respondents also have the same results, namely there is a positive relationship between attitudes and intention to choose president candidate. The significant result in the relationship between attitude and intention to choose president candidate is similar to previous study in the same context.
Although this study aims to find out the differences in results between students who have a background in science and nonscience, but it turns out that the results of this study show no difference. This can be caused by several factors. First, the context of this research is related to the presidential election. With only two candidate candidates who have fought in the same election in 2014, the respondents have had an evaluation of the two presidential candidates. A period of 5 years since the two fought in 2014 can indeed influence the choice of voters while still choosing the choices that were formerly or will be different in 2019. However, in the last 5 years, voters already have knowledge of these candidates. Other factors can be caused by respondents in this study coming from a university based on Christianity. That is, respondents in this study (where the results of the study also showed) tend to choose only one candidate, namely Jokowi who is perceived as a leader who provides freedom for minority religions in Indonesia. Thus, it can be said that although there are differences in character between science and non-science students, but because the context of this research is that the presidential election which only gives two presidential candidates makes all the differences between voters and science and non-science backgrounds make the difference seems invisible. There is no visible difference in the background of voters then it is also seen nationally where with a large diversity of cultures, backgrounds, religions, and others in Indonesia, it will continue to converge on only one choice in 2019.
The results show that there are no differences in results and three of the eight hypotheses are supported for both types of samples. Charm is the only image factor that has a positive relationship with the attitude of choosing candidates. Next, the results confirm the relationship between attitudes and intentions to choose candidates.

Managerial Implications
This research has at least 2 managerial implications. First, Jokowi's political communication must continue to be maintained based on politeness, simplicity, humility, and other things which are then reflected in Jokowi's aura and positive charm. Maintaining a positive charm can take the heart of the community. It is realized that although in many societies including the Indonesian people as a rational society, it cannot be separated from matters relating to the heart/feeling. Secondly, Jokowi's charm will influence the attitudes of voters. Specifically, good charm will influence voters to have a positive attitude. This positive attitude will then shape one's intention to elect a presidential candidate in the coming year. Therefore, maintaining Jokowi's charm is part of the effort to shape and maintain a positive attitude towards Jokowi so that later it can be reflected in the people's choice toward him as president in 2019.

Limitations and Direction for Future Research
This research cannot be separated from several limitations. First, this research uses non-probability sampling by selecting student respondents purposively at one university. Therefore, the results do not reflect the generalization of all young voters in Indonesia. Subsequent research can replicate this research so that the same or different results can provide an understanding of the effect of the image on the intention of choosing the president with an attitude as an intervening variable. Second, the data for this study were collected in just one period. In other words, this research applied cross-sectional study. Therefore, causality is not specified in this study. Future research can try to establish a causal relationship between charm and attitude.