Factors Affecting the Decision Using Theory of Planned Behavior (Case Study: Election of the 2019 President)

Indonesia has adopted a democratic system since the 1998 reform era which includes many aspects, including politics. The Indonesian people are the highest authoritys holder who are entitled to the opportunity and voice in regulating government policies. With different racial, religious, ethnic backgrounds, it is hoped that all indonesians citizen can participate in this election. Although the distribution of the Indonesian population is concentrated in Java, it does not make the focus of political marketing only on Java. There are restrictions on age, education, geographical or socio-economic in political marketing, then the right marketing strategy is needed. Political actors need to know the characteristics and behavior of voters to determine an effective communication strategy. The purpose of this study is to analyze the factors that influence decisions outside of Java using Theory of Planned Behavior. Data were processed using SEM analysis with AMOS 23.0. The results showed that the factors that influence the intention to vote outside of Java are attitudes and subjective norms. 10% of the total respondents stated that they had not made a choice when the questionnaires were distributed, so there is a possibility that swing voters will determine the votes in the last seconds of the election.


Introduction
Indonesia has adhere a democratic system since the reform era in 1998 (Monje Amor et al., 2019). Democracy allows all citizens to participate in the formulation, development and decision-making that affect their lives. Therefore, it can be concluded that the Indonesian people are the highest power holders in Indonesia who have the right to have the opportunity and to have a voice in regulating government policy. Democracy in Indonesia includes many aspects, one of which is politics. In 2004, for the first time in Indonesia's history the people directly elected the president and vice president. This is in accordance with the third amendment to the 1945 Constitution Article 6A paragraph (1) stipulates that "the president and vice president are directly elected as a pair by the people".
The President of the Republic of Indonesia holds the power as both head of state and head of government (Wahyono, 2022). This is based on the results of the amendment to the 1945 Constitution Article 4 Paragraph 1 which reads, "the president of the Republic of Indonesia holds governmental power according to the Constitution." As head of state, the president has political rights which are determined in accordance with the constitution of a country. Meanwhile, in carrying out his duties as head of government, the president needs a foundation or basis as a guide in running the government of a country.
Presidential and vice President elections are held every five years starting from 2004,2009,2014 and held again in April 2019 with two pairs of presidential and vice presidential candidates. According to (Tabroni, 2014) the intense political competition encourages every candidate who is a party icon to make unique and interesting creations in order to get public attention.
The use of the concept of political marketing carried out by parties or political figures is an effort to be further recognized by the public and to influence people's attitudes in making political choices (Alie, 2013).
The 2019 election is the most appropriate time for candidates to do their own marketing, as stated by (Alie, 2013) political marketing is a marketing concept approach that is commonly applied to products and services as an approach applied to political science. Business marketing and political marketing have the same process of activity, only the means of transaction and the product being sold differentiate them. In business marketing, a transaction occurs because of the communication made by the seller so that consumers are interested in the products and services offered. Transactions in political marketing take the form of voices, with products and services being traded in the form of promises and favors and information communicated to voters.
Each candidate competes in attracting sympathy through the political marketing they offer to the public. Research by (Axhami et al., 2015) in Albania revealed that the factors that influence decision to choose are political products, political advertising, public relations and direct marketing resulting in voting behavior.
Indonesia is known as a complex democracy based on its size, geography and election administration. Consisting of several islands with different racial, religious and ethnic backgrounds, it is hoped that the entire population of Indonesia can participate in this election. Even though the distribution of Indonesia's population is still concentrated in Java, this does not make the focus of political marketing only centered on the island of Java. There are restrictions on age, education, geographic area, or socio-economic conditions in political marketing not to be chosen and focus on one of them alone, but to determine the strategy for each segment so that it can be reached and intervened (Tabroni, 2014).
The 2019 Presidential Election is the right moment to identify the characteristics of voters and the factors that most dominate the decision to elect a presidential candidate. Therefore, researchers are interested in analyzing the factors that influence decisions using Theory of Planned Behavior (a case study of the 2019 presidential election outside Java). Theory of planned behavior (TPB) is an improvement from reasoned action theory. Reasoned action theory has scientific evidence that the intention to carry out certain actions is caused by two reasons, namely subjective norms and attitudes toward behavior (Ajzen & Fishbein, 1975). Theory of Reasoned Action has limited inability to handle individual behavior control. So that several years later, (Ajzen, 1988) added one factor, namely perceived behavioral control. It can be concluded that intention precedes a behavior represented by three beliefs, namely behavioral beliefs, normative beliefs and control beliefs that influence each other, where these are related to the construction of attitudes, subjective norms, and behavioral control. perceived behavior control.
According to (Soleimanpour Omran, 2014) attitude is a brief evaluation of everything based on cognitive, emotional and behavioral information. Another definition explains that attitudes are evaluative statements or judgments relating to an object, person or event (Weiss, 2002). (Shook & Bratianu, 2010) argue that a person's attitude towards something depends on the result, the more positive the result, the better a person's perception and the higher the intention to perform the behavior.
Meanwhile, the definition of subjective norm according to (Huda et al., 2012) is a person's perception or assumptions about other people's expectations, certain behaviors that someone will or will not do. The reason for the direct effect of subjective norms on intention is that people can choose to carry out a behavior, even though they themselves dislike the behavior or its consequences (Listyoningrum & Albari, 2017). Another definition of (Anggraini & Walyoto, 2018) Subjective norms are individual perceptions of environmental pressures in their lives regarding whether certain behaviors are carried out or not.
According to (Ajzen & Madden, 1986) behavior control refers to a person's perception of his ability to display certain behaviors. Behavior control is an individual skill in reading the situation themselves and their environment (Ghufron & Rini Risnawita, 2010). In line with (Aida et al., 2019) opinion that someone will have the intention to carry out a behavior when they have the perception that the behavior is easy to show or do, because there are things that support the behavior. In general, perceived behavior control is a view of the ease and difficulty of showing attitudes.
Behavioral intention (intention) is determined by attitudes, subjective norms and conscious behavior control (Albery & Munafo, 2011). A person develops an intention to change something, with a desire, the desired behavior must be planned, initiated, and maintained (Sniehotta et al., 2005). An indication of one's readiness (Ajzen, 2005) defines intention as an arrangement of actions in which if there is a suitable time and opportunity it will be realized in the form of action.
Decisions are the result of a reasoned process in which behavior is influenced by perceived attitudes, norms, and behavioral control (Phan & Zhou, 2014). According to (Schiffman & Kanuk, 2013) a decision is the selection of an action from two or more alternative options.
In particular, this study aims to identify the characteristics of voters outside Java in the 2019 presidential election, and identify the factors that influence the election decisions of the 2019 presidential and vice-presidential candidates. outside Java Island and become an input for candidates in order to get maximum votes in the next election.

Research Method
The data used in this study are secondary data. Secondary data is data collected by other people for different purposes from the research objectives it formulates (Sumarwan et al., 2013). Secondary data can be obtained from previous research, government agencies, books, literature, theses, theses, journals, archives and the internet. Meanwhile, the sample in this study is people outside Java who are at least 17 years old and have registered as permanent voters who use their voting rights in the 2019 election and use non-probability techniques. In this study, using (Didin, 2019) with a total sample size of 343. Sampling used the Quota Sampling Technique which was divided based on the number of permanent voter lists in each province outside Java which was released by the KPU.
The research went through a two-stage analysis process. The first analysis is descriptive analysis which is used to analyze the characteristics of voters. Furthermore, the second analysis is carried out, namely CB (Covariance Based) SEM analysis using the AMOS 23.0 application to determine the factors that influence the decision to choose. Both analyzes will produce managerial implications that serve as evaluations of presidential candidates as well as recommendations for increasing the effectiveness of political marketing.
The methodology is based on Theory of Planned Behavior (TPB) by including factors as background. According to (Ajzen, 2015), a person's behavior is influenced by three components, namely attitudes, subjective norms, and perceived behavior control. The next research step is to prove the perceived influence of the TPB on the intention to elect candidates in the election.
This study has twenty-five indicators, three independent variables and two dependent variables. The independent variable consists of intention, subjective norm and perceived behavior control which is connected to the dependent variable, namely the intention to choose which is then associated with the decision to choose. Figure 1 below is a theoretical model based on the theory of planned behavior approach which refers to the (Ajzen, 2015).

Figure 1. Research Model
From the picture above, the hypothesis in this study can be described as follows: H1: The attitude variable (X1) affects the intention to choose (Y). H2: Subjective norm variable (X2) affects the intention to choose (Y) H3: Perceived behavior control variable (X3) affects the intention to choose (Y) H4: The intention variable (Y) affects the decision to choose a candidate (Z)

Results and Discussion
There are two pairs of presidential candidates and vice presidential candidates in the 2019 presidential election with their respective vision and mission. The presidential election process begins with a campaign on 23 September 2018 -13 April 2019, a quiet period on 14-

A. Descriptive Analysis 1. Characteristics of Voters
The characteristics of respondents in this study were divided based on domicile according to the Identity Card (KTP), residence domicile, gender, age (years), marital status, religion, latest education level, employment status, income level / month, and frequency of participation in the presidential election. While the characteristics of the most dominant respondents in this study can be seen in Table 1.  (2020) Based on the data in Table 6, it can be seen that the largest number of respondents with a composition of 12.24% of all respondents came from North Sumatra, Lampung in second place and South Sulawesi in third with a proportion of 7.58% with a resident domicile. 76.97% come from urban areas and the rest come from rural areas. Female respondents in this study were as many as55.10% and as many male respondents 44.90%. The distribution of questionnaires conducted online made the age distribution 17-25 years dominating by proportion52.48% and the second position is 26-35 years old by proportion 26.24%. Due to the relatively young age of vulnerability, this is in accordance with the respondent's marital status63.56% unmarried and 34.99% are currently married.
The majority of Indonesian people are Muslim, this makes Muslim respondents in the first position in this study with a percentage of 83.67% and the second percentage 10.50% of respondents are Protestants. 59.77% of respondents in this study had completed S1 / S2 / S3 education and 32.36% have completed high school education / equivalent. In this research there are40.82% of respondents who are students, 20.12% private employees and 16.03% Government employees(ASN). The dominance of the occupational profession in this study is student / student with a total percentage39.07% followed by lecturers / teachers of 19.24% and in the third position is the office staff of 12.24%, with the most dominant respondent's income level is ≤ Rp. 1,500,000 with a percentage33.24% and 20.99% earn Rp. 2,500,001 -Rp. 5,000,000.
One of the criteria for respondents in this study is that they have participated in the presidential election at least once. And this is in accordance with the number of respondents who dominate, namely equal to35.57% has a frequency> 2 times and 28.28% of respondents have participated in the presidential election once.

Model Specifications
This initial research model is formulated based on a theory or previous research. The specification of the research model, which represents the problem under study, is important in SEM.

Figure 2. Initial Research Model
The initial research model can be seen in Figure 2, it can be seen that the different codes used for each variable and its indicators. This code makes it easier for researchers to process the SEM AMOS 23.0 application. Each code used is described in Table 3 below.

Measurement Model
This evaluation is carried out on each construct or measurement model (the relationship between the latent variable and the observed variable) separately through the validity and reliability of the measurement model. Measurement of the validity of the SEM model in this study uses First Order Confirmatory Factor Analysis (First Order CFA), where a variable is said to have good validity for the construct or latent variable if the standardized loading factor value is greater than or equal to the value. critical value is 0.50 (Igbaria et al., 1997) or the t value of the standardized loading factor is greater than or equal to the critical value of 1.96 (Ghozali, 2014): Where standardized loading can be obtained directly through the output of the AMOS program application, and εj is a measurement error for each indicator or observed variable (Fornell & Larcker, 1981). The cut-off level for being able to say that construct reliability is good is greater than 0.60, while the cut-off level for being able to say that the average variance extracted is good is greater than 0.50 (Ghozali, 2014).
Based on the output of the AMOS calculation results, the standard factor load values are obtained and then used to calculate the value of the construct reliability coefficient which is summarized in the following figure:

Figure 3 Estimation Results of Standardized Loading Factors
Based on Figure 3 above, it can be seen that there are standardized loading factor values that have a standard load factor value that is smaller than 0.5, namely NS 5 , NS 6 , KPP 1 , KPP 2 , KPP 3 , NM 1 , NM 2 , and KM1. If there is a standard factor load value that is smaller than the critical value, then the associated unobserved variable can be eliminated from the model. However, if the standard factor load value is still ≥ 0.3 then the related variable can still be considered for deletion (Igbaria et al., 1997). Because there is a standard factor load value of observed variables less than the critical value of 0.3, namely NS 5 , NS 6, KPP 2 , NM 1 , and KM 1 , the observed variables NS 5 , NS 6 , KPP 2 , NM 1 , and KM1 are removed from the model. The following shows the estimation

Figure 4. Estimation Results of Standardized Loading Factors
Based on the picture above, it can be seen that there is a standardized loading factor value that has a standard load factor value that is smaller than 0.5, namely KPP1 and KM2. If there is a standard factor load value that is smaller than the critical value, then the associated unobserved variable can be eliminated from the model. However, if the standard factor load value is still ≥ 0.3 then the related variable can still be considered for deletion (Igbaria et al., 1997). Because there is no standard factor load value of observed variables less than the critical value of 0.3, the observed variables KPP1 and KM2 are not removed from the model.

Structural Model
This section deals with the evaluation of coefficients or parameters that indicate a causal relationship or the effect of one latent variable on another latent variable. In summary, the results of the calculation of the coefficients are presented in the following table:

C. Hypothesis test
Hypothesis testing is done to measure the causality relationship between variables in the SEM model in this study. Significant testing criteria in SEM are based on a critical point value of 1.96 where the t-value is greater than or equal to the critical point (t-value ≥ 1.96) indicates that the parameter value is statistically significant. The next column contains the estimation results of standardized coefficients between one latent variable and another.The following shows the results of hypothesis testing for each hypothesis.

Effect of Attitude (X1) on Intention to Choose Candidates (Y)
Statistical hypothesis for Hypothesis 1: H0: Attitude (X1) does not have a significant effect on Candidate Voting Intention (Y) H1: Attitude (X1) has a significant effect on the Intention to Choose Candidates (Y) Furthermore, based on the above hypothesis which was carried out using AMOS software, the following values were obtained:  (2020) From the results of table 4, the path coefficient value of 0.352 shows that the direction of the relationship between Attitude (X1) and Intention to Choose Candidates (Y) is positive or unidirectional, meaning that if Attitude (X1) increases, the Intention to Choose Candidates (Y) will increase, as well as otherwise. The relationship between Attitude (X1) and Intention to Choose Candidates (Y) is significant in the 2-tailed test (t table = 1.96) with the t-count value of 4.332 greater than the t table. Thus, H1 is accepted, meaning that Attitude (X1) has a significant effect on the Intention to Choose Candidates (Y).

Effect of Subjective Norms (X2) on Intention to Choose Candidates (Y)
Statistical hypothesis for Hypothesis 2: H0: Subjective Norms (X2) have no significant effect on Candidate Voting Intention (Y) H2: Subjective Norms (X2) have a significant effect on the Intention to Choose Candidates (Y) Furthermore, based on the above hypothesis which was carried out using AMOS software, the following values were obtained: Furthermore, based on the above hypothesis which was carried out using AMOS software, the following values were obtained:  (2020) From the results of table 7, it is obtained that the path coefficient value is 0.064, indicating that the direction of the relationship between the Intention to Choose a Candidate (Y) and the Decision to Choose a Candidate (Z) is positive or unidirectional. The relationship between Intention to Choose Candidates (Y) and Decision to Choose Candidates (Z) is significant in the 2-tailed test (t table = 1.96) with the t-count value of 1.001 smaller than the t table. Thus, H4 is rejected, meaning that the intention to choose a candidate (Y) does not significantly influence the decision to choose a candidate (Z).

Conclusion
Attitude variables and subjective norms have an influence on intention to choose. While the perceived behavior control variable does not have a significant effect on the intention to choose. intention to choose does not have a significant effect on the decision to choose. The existence of news and information obtained as well as situations and conditions that can influence respondents in making the final decision in the presidential election are one of the obstacles in the formation of loyalty from voters to candidates. In addition, 10% of the total respondents also stated that they had not made a choice when the questionnaires were distributed. So that the possibility of swing voters, namely voters who determine the votes in the last seconds of the election. The existence of news and information obtained as well as situations and conditions that can influence respondents in making the final decision in the presidential election are one of the obstacles in the formation of loyalty from voters to candidates. In addition, 10% of the total respondents also stated that they had not made a choice when the questionnaires were distributed. So that the possibility of swing voters, namely voters who determine the votes in the last seconds of the election. The existence of news and information obtained as well as situations and conditions that can influence respondents in making the final decision in the presidential election are one of the obstacles in the formation of loyalty from voters to candidates. In addition, 10% of the total respondents also stated that they had not made a choice when the questionnaires were distributed. So that the possibility of swing voters, namely voters who determine the votes in the last seconds of the election.
For the next presidential candidate, it is advisable to maximize the formulation of political marketing strategies by considering factors from the research results that have been tested significantly, namely subjective attitudes and norms. It is necessary to take an effective approach so that voters can be confident in determining their hearts to vote according to the conditions and situations they are experiencing. Meanwhile, future researchers are advised to investigate more deeply about the existence of swing voters. And it is hoped that it will be able to reach those who do not have internet access as well as adding other factors that are in accordance with the development of environmental conditions in Indonesia so that more accurate and up-to-date information can be obtained.