Vote-Selling as Unethical Behavior: Effects of Voter’s Inhibitory Self-Control, Decision Toward Vote-Buying Money, and Candidate’s Quality in Indonesia Election

This study examines two stages of vote-selling among voters: to accept or refuse money, and then to vote for the vote-buyer or competitor. Using unethical behavior and money-as-temptation as a framework, we predict that the amount of the money will influence the decision to accept, but that the effect will be lower among voters with higher inhibitory self-control (ISC). We also predict that accepting money will influence the vote choice, but that the influence will be lower when the competitor has higher levels of integrity and leadership. Overall, the voters decision on money offered will moderate the relations between amount of money and vote choice. A population-based survey experiment with 1,220 participants of Indonesian voters was conducted to test these predictions. Participants were randomly assigned to one of 12 groups using a 4 (money offered: IDR50,000 vs. 100,000 vs. 150,000 vs. none) × 3 (candidate's personal quality: high leadership competitor vs. high integrity competitor vs. equal) between-group design. They made two hypothetical decisions: to accept or refuse the money, and to vote for the vote buyer or the competitor. The results showed that participants were signicantly tempted by larger amount of money, but the effect was moderated by ISC. However, there were no significant interaction between decision on money offered and candidates’ quality in influencing vote choice. Results from mediation analysis showed that accepting money was significantly mediate the influence of money on vote choice. Theoretical implications for understanding the practice of vote-buying and vote-selling are discussed.

In vote buying, voters have to make at least two decisions. First, they have to decide whether to accept or refuse the money offered. Second, they have to decide which candidate to vote, i.e., the vote buyer or another candidate. Past research on vote-buying has found that socio-psychological and political factors, as well as voters' characteristics, predicted voters' acceptance and experience in vote-buying-factors such as reciprocity, party identification, political attitudes, broker's social network, and other demographic variables (Aspinall, 2014;Brusco et al., 2004;Finan & Schechter, 2012;Jensen & Justesen, 2014;Muhtadi, 2019;Vicente, 2014). However, this research has not specifically focused on the decision voters made when they were being offered material inducements and how those decisions affect their vote choice from the perspective of unethical behavior. For Indonesia, focusing on voters' decisions is vital because voters still have agency in voting. On one side, voters are depicted as easily persuaded by money (Fionna, 2015). But on the other side, previous research has found that voters tend to betray or vote for another candidate, rather than the one who bought their vote (Aspinall, 2014;Fariz, 2014). That is why understanding what psychological factors affect voters' decisions are essential in explaining vote-selling in an election context. This article aims to contribute to the understanding of vote-selling decisions made by voters in two stages of decisions and seeing them as a form of unethical behavior, testing the effect of money, inhibitory self-control, and candidates' quality on voters' decisions. With this perspective, voters' decisions to sell their votes are considered unethical, since it is a form of acceptance of bribery. Past research on unethical behavior showed that individual differ ences, as well as social-psychological or situational factors, might influence an unethical decision, such as personality, self-control, money or wealth, and social identity (Gino et al., 2011;Gino, Ayal, & Ariely, 2009;Vohs, Mead, & Goode, 2008). As a form of unethical behavior, voters' decisions to sell their votes can be influenced by individual and social or situational factors. In the first stage, this study tests whether the scale of money serves as a temptation for voters to accept the bribery, in which self-control can play a crucial role (Hicken, Leider, Ravanilla, & Yang, 2014). In response to the money offered, voters have to decide whether to accept or refuse it. To refuse the offer, they need to exercise inhibitory self-control (de Ridder, de Boer, Lugtig, Bakker, & van Hooft, 2011), which is to be tested as a moderating variable between the amount of money offered and the response to the offer. We also test whether decision to accept will mediate the relations between the amount of money offered and vote choice.
In the second stage, this study tests the candidate personal quality as situational factors affecting vote choice in the ballots. In Indonesia elections, candidates' personal quality was related to voting choice (Liddle & Mujani, 2007), where voters developed a social impression of certain candidates and based their choice on it. Our study aims to test this proposition in the context of vote-buying. Voters choose between the candidate who offered money and one who did not. In this stage, the candidate's leadership quality (Bean & Mughan, 1989;Liddle & Mujani, 2007;Miller & Shanks, 1996) and integrity (Liddle & Mujani, 2007;Mujani et al., 2012Mujani et al., , 2018 played an important role as situational factors affecting vote choice. Previous studies found that different candidate's background could vary their strategy in bribing voters (Collier & Vicente, 2012;Conroy-Krutz & Logan, 2012;Vicente & Wantchekon 2009). However, what kind of candidate could win voters' decisions other than the vote-buyer candidate and how much bias was caused by receiving the money remained open questions that need to be answered. The present study tests candidate quality as a potential moderator between the decision on the financial offer and the vote.
We apply a hypothetical vote-buying scenario with specific information that we want to test, minimizing other confounding variables that could interfere if we apply the real situation of actual election. In this case, we follow Lau and Redlawsk's (2006) design which studied information processing during an election campaign, in which they presented vote choice alternatives and the information which is available to the voters. We present the result of hierarchical logistic regression analysis and mediation analysis, and discuss our findings and limitations, and also the implication of our results.

Vote-Buying and Selling as Unethical Decision
Vote buying has been described as occurring at "the moment an inducement is offered by a candidate or a candidate's agent with the clear intention of harvesting the recipient's vote" (Transparency International, 2004, p. 77). Another definition refers to it as "dispensing cash to individual voters as an incentive to cast their ballots for a particular candidate" (Aspinall, 2014, p. 104). Both definitions stress the purpose is to bribe voters and persuade them to sell their votes.
Vote-buying and selling is an illegal practice, including in Indonesia. It harbors negative consequences for the strength of democracy because it violates the equality principle (Stokes, 2007) and reduces the quality of public policy and public service produced after the election (Desposato, 2007). For those reasons, democracy activists oppose the vote-buying practice (Ocantos, de Jonge, & Nickerson, 2013).
Research on vote-selling has often sought to understand whether receiving a monetary offer affects people's actual voting decision by asking whether voters are being offered any inducement (i.e., money) before election day and then asking about their vote choice on the election (e.g., Bratton, 2008;Brusco, Nazareno, & Stokes, 2004;Conroy-Krutz & Logan, 2012;Weghorst & Lindberg, 2013). Researchers have compared those two results to measure how much influence the inducement has on vote choice. Meanwhile, the decision of whether to accept the money has remained un derstudied. Other studies have applied field-based experimental designs in conjunction with campaigns that denounce vote-buying, seeking to educate voters and persuade them to refuse the money offered (Hicken et al., 2014;Vicente, 2014). While those studies have informed us of the possibility of voters' responses to the money offered, they have not given us a clear explanation of voters' decision-making process when they receive the offer (i.e., accept or refuse).
As a form of bribery, even considered electoral corruption (Lehoucq, 2007), vote-buying and selling can be consid ered a test of voters' ethical behavior. In the face of money offered, voters can either stand for ethical standards reflected by refusing the money or can relax their ethical standards by accepting the money.
Unethical decisions are defined as a decision that is "either illegally or morally unacceptable to the larger communi ty" (Jones, 1991, p. 367). An unethical decision or unethical behavior is characterized by a violation of widely accepted (societal) or moral norms (Rest, 1986). Research on unethical decisions or behavior reveals many forms of behaviors that are considered unethical, such as lying, cheating, stealing, and other behaviors that can be considered a violation of ethical norms. The concept was developed from social psychology and moral philosophy, but then applied in organizations and marketing behavior.
Previous research has shown that individual and situational factors can influence unethical behavior. Individual differences that are related to unethical behavior are Machiavellianism and locus of control (Kish-Gephart, Harrison, & Treviño, 2010). While situational factors influence individuals' unethical decisions, such as the norms and culture (Treviño, Butterfield, & McCabe, 1998), the code of conduct (McCabe, Treviño, & Butterfield, 1996), and the reward system (Hegarty & Sims, 1978). Other immediate situational factors in psychological research have found that the visual stimulus of money (Vohs, Mead, & Goode, 2008) and the visibility of abundant money could lower one's ethical standard (Gino & Pierce, 2009). Those research findings indicate that people may be influenced by their differences and also social or situational surrounding in arriving at unethical behavior.
Previous research has been very limited in addressing unethical behavior in political behavior, especially during the election, in the form of vote-buying and selling. During the election campaign, vote buyers approach the voters to sell their votes. They offer many kinds of material inducements, with money as the most widely used.
The money offered in vote-buying poses a temptation for voters to sell their votes (Hicken et al., 2014). In Hicken et al. (2014) study in the Philippines, the researchers found that voters' self-control could be partially induced by encouraging voters to refuse the money in the first place. However, the research did not study the self-control of the individual as a factor which might explain voters' decisions in the first stage of vote-buying.
Since vote-buying is seen as a common social phenomenon during the election in Indonesia (Aspinall & Sukmajati, 2015;Djani & Vermonte, 2014), where law enforcement for the election violation is also unreliable (Anggraini, 2016;Fariz, 2014), even involving the election committee (Najib, 2015;Ridwan, 2015) and many candidates and their teams use norms and culture to conceal the unethical vote-buying transaction (Paskarina, 2015), more attention is needed to study the self-control as an individual trait that makes voters able or not able to refrain from selling their vote in this unethical transaction.

Stage I: Voters' Acceptance or Refusal of Money and the Role of Inhibitory Self-Control
The role of individual differences in the propensity to engage in unethical behavior has been found in previous studies. Hegarty and Sims (1978) found that locus of control, economic and political value orientation, and Machiavelianism sig nificantly covaries with unethical behavior in marketing decision making. While individual factors related to controlling behavior in restraining unproductive behavior, such as self-control, were rarely mentioned in explaining the ability to resist the money offered in political behavior studies such as vote-buying.
Self-control is the ability to override or change one's inner responses, interrupt undesired behavioral tendencies, and refrain from acting on them (Carver & Scheier, 1998;Tangney, Baumeister, & Boone, 2004). It is the capacity owned by individuals to change and adapt the self to produce a compatible relation between self and environment (Ent, Baumeister, & Tice, 2015;Rothbaum, Weisz, & Snyder, 1982). Self-control is reflected in behaviors such as resisting temptation, breaking bad habits, and having self-discipline (Tangney et al., 2004).
There are two types of self-control, namely initiatory and inhibitory self-control. Initiatory self-control refers to an individual's capability to perform goal-oriented behaviors (de Ridder et al., 2011). While inhibitory self-control entails the ability to replace or change one's internal responses, such as stopping certain unwanted behavioral tendencies (Tangney et al., 2004).
In vote-selling, the money offered by vote buyers acts as a temptation for voters (Hicken et al., 2014). A temptation creates a desire that conflicts with another value or objective and must be either satisfied or restrained (Hofmann, Baumeister, Förster, & Vohs, 2012). For the voters, accepting the money means that they have participated in an illegal and unethical transaction. To refuse the money, voters need inhibitory self-control.
Thus far, money in vote-selling was only mentioned as one of the resources used by vote buyers (Aspinall, 2014;Aspinall & Sukmajati, 2015;IFES, 2014;Vicente, 2014), as if it had the same value as other offerings in the form of goods, such as household utensils or prayer equipment. However, research about resources showed that money was different from other resources and could not be exchanged with other resources that were considered inappropriate to be exchanged with money, such as love (Foa & Foa, 1976, 2012. Moreover, the presence of money (Vohs, Mead, & Goode, 2008) and the visibility of abundant money could trigger unethical behavior (Gino & Pierce, 2009). This research indicates that money cannot be considered the same as other forms of goods in vote-selling, but money may have a unique psychological effect in influencing voters' into more unethical behavior.
In vote-selling, eventhough the money offered is suffused with negative connotations as a violation of the law, undemocratic, a form of bribery, and a taboo transaction (Lehoucq, 2007;Stokes, 2007;Tetlock, 2003;Tetlock, Kristel, Elson, & Lerner, 2000), a larger amount of money will pose a higher temptation for the voters to accept it.
In response to this temptation, inhibitory self-control, or the ability to restrain oneself from undesirable behaviors (de Ridder et al., 2011), might determine whether the individual resists or succumbs. The temptation posed by the money may be restrained by voters' inhibitory self-control. Based on this reasoning, we predict the voters' decisions based on money and inhibitory self-control in stage one as follows: H1: Participants' decisions toward the money offered will be based on its amount: the more money is offered the more likely it will be accepted.
H2: The effect of money offered on participants' decision to accept will be lower among voters with higher inhibitory self-control.

Stage II: Decision Towards Money Offered and Candidates' Quality in Vote Choice
Vote choice can be influenced by voters' decisions toward the amount of money offered, but other factors might also interact. In Indonesia, Muhtadi (2019) found that vote-buying only affecs a small portion of voters. Other socio-political variables that might influence vote choice include candidates' campaign teams surveillance at the ballot location or the coercive strategy (Bratton, 2008;Collier & Vicente, 2012;Vicente & Wantchekon, 2009) or personal network (Fauzan, 2009). In Indonesia, factors that related to vote choice in vote-buying includes campaign network, clientelism, and other issues related to the election (Aspinall & Sukmajati, 2015;Damanik, 2015;Jafar, 2015;Muhtadi, 2019). All of those factors shows efforts to ensure that targeted voters receive the money offered, hoping that voters will pay it back by selling their vote, or voting for the vote-buyer candidate (Muhtadi, 2019;Rahman, 2015). However, voters' consideration in deciding their vote choice under the direct, free, and secret election like in Indonesia is still unexplored.
One issue that has been a concern in vote choice and vote-buying is that voters will vote for candidates with bad qualities because of the money, while other candidates who have good qualities remain unelected, hindering democracy and policymaking (Khemani, 2015). Despite this concern, there has been limited effort to understand the mechanism of vote-selling where at least two candidates have different qualities. So far, existing research has explored differences between incumbent and challenger in how they employ different modes of persuading voters using material inducements (Collier & Vicente, 2012;Conroy-Krutz & Logan, 2012;Vicente & Wantchekon, 2009), or in the scope of election (Hicken et al., 2014).
In Indonesia, the vote choice is also highly influenced by the candidates' personal qualities. For example, during the 2004 Indonesia Presidential election, Yudhoyono won the election because voters saw him as superior in personality compared to other candidates who were nominated by big parties, such as Megawati Soekarnoputri, an incumbent and Head of Indonesia Democratic Party of Struggle (PDIP), and Wiranto whom Golkar Party nominated (see Liddle & Mujani, 2007). Two aspects of candidates' qualities that relate to voting choice were leadership and integrity (Liddle & Mujani, 2007;Mujani & Liddle, 2010;Mujani et al., 2012Mujani et al., , 2018. Election research on political leadership identifies several personal qualities in leaders, such as competence, effectiveness, integrity, empathy, and likeability (Bean & Mughan, 1989;Liddle & Mujani, 2007;Miller & Shanks, 1996). Together, those qualities form voters' impressions of a candidate's leadership potential. However, among those qualities, integrity has been the one that Indonesian voters think is most important (Liddle & Mujani, 2007). Integrity refers to honesty and reliability, and is an important quality that relates to anti-corruption, and therefore unethical behavior. It is different from other personal qualities, such as competence which refers to the ability to lead (Liddle & Mujani, 2007). For this reason, we separate leadership and integrity.
This influence of personal qualities is particularly great when the electoral mechanism provides an opportunity for candidates to appear as individuals -not only representing a political party -and when the differences between party programs and ideologies are relatively obscure (Barisione, 2009;Lobo, 2008;Stewart & Clarke, 1992). Those characteristics of a political situation can be observed in Indonesia's multi-party system and fragmented political party ideology.
This importance of candidates' quality in determining vote choice had not yet been tested in understanding vote choice in a vote-buying situation. Voters can vote freely in the vote-buying context in Indonesia, whether they accept or refuse money offered by vote-buyer candidates. Money may influence vote choice, but not always (Hicken et al., 2014;Muhtadi, 2019). Because candidates' personal quality has been confirmed to influence vote choice, we will test its effect on vote choice in a vote-buying situation.
H3: Candidates' personal quality in leadership and integrity will moderate the effect of voters' decisions toward money offered on vote choice. Voters who accept the money offered will not vote for the vote-buyer candidate when the other candidate has better leadership or integrity.

Accepting Money as a Mediator Between Money and the Vote Choice
The question about vote choice in vote-selling has been posed by researchers in previous studies. Do voters sell their votes in exchange for some amount of money? If yes, what makes them decide the way they do?
Previous research reveals contradicting results. In Indonesia, the money offered was not necessarily followed by vote choice (Muhtadi, 2019;Rahman, 2015). Candidates and their campaign teams who reported practising vote-buying often faced the reality that the money they offered during the campaign did not automatically bring them to victory (see Muhtadi, 2019;Rahman, 2015). However, there have also been empirical research findings that reveal that voters were also persuaded by money (Fionna, 2015).
Many factors are related to voting choice in vote-buying, such as network campaigns, clientelism, and other issues related to elections (Aspinall & Sukmajati, 2015;Damanik, 2015;Jafar, 2015;Muhtadi, 2019). However, previous research also showed that all efforts were made to ensure that targeted voters receive the money offered, hoping that voters will pay it back by selling their vote, or voting for the vote-buyer candidate (Muhtadi, 2019;Rahman, 2015).
Money has been one of the resources thatwas widely used in vote-buying and vote-selling (Nichter, 2014). Money was also believed to be most effective in ensuring voters' 'loyalty' to the vote buyer (Aswad, 2015;Hamdi, 2015;Rahman, 2015). In a setting where vote-buying has been a common practice, such as in Indonesia, the amount of money offered also matters in securing votes. For example, in Madiun, East Java, a legislative candidate who gave a higher amount of money to the voters successfully won (Hamdi, 2015). The same pattern also happened in North Jayapura, Papua (Ridwan, 2015), where larger sums of money determined the outcome of the election.
Despite the case that confirmed the influence of the amount of money on vote choice (e.g. Hamdi, 2015;Ridwan, 2015), other research found that accepting or refusing vote-buying money could influence vote choice. When voters were given an anti-vote buying campaign that suggested them to 'not accept money', voters would be more likely to vote for their preferred candidate, compared to when they received the campaign to 'take the money but vote for their own preferred candidate (Hicken et al., 2014). A similar experimental field result with an anti-vote buying message was also conducted in West Africa (Vicente & Wantchekon, 2009). Those findings showed that voters' acceptance or refusal of money offered to buy their votes was crucial in determining vote choice. Based on those previous research findings, we formulate our fourth hypothesis that predicts the mechanism of how the money will influence vote choice: H4: Participants' decision toward money offered will mediate the relations between the amount of money offered and vote choice: Those who accept the money will tend to vote for the vote-buyer candidate and vice versa.

Method Sample and Design
We applied population-based survey experiment (Mutz, 2011). We selected 1,220 voters from 34 Indonesian provinces using a stratified-clustered random sampling technique 1 (see Supplementary Material 1 for the number of participants from each province). The sampling error under the assumption of simple random sampling is ± 2.87%. The data collection was conducted by an Indonesian political research company as a part of a national survey.
Selected participants were interviewed face-to-face by trained interviewers. Participants were randomly assigned to one of 12 groups according to a 4 (amount of money offered: IDR 50,000, 100,000, 150,000, or none) x 3 (candidate's personal qualities: competitor not offering money is better in leadership qualities, better in integrity, or equal) between-1) The population of Indonesian voters in 34 provinces is stratified based on province, rural-urban, and gender. In each province, we randomly selected villages (the smallest administrative area) or kelurahan (urban village) as the primary sampling unit (PSU) applying probability proportional to size. In each selected village, we listed all the Rukun Tetangga (RT, the smallest neighborhood unit) and randomly selected 5 RT. In each selected RT, we listed all the household and randomly selected two hoseholds. And in each selected household, all of the household members aged 17 years old and older were listed and randomly selected for one female or male respondent with Kish Grid. The selected participant then interviewed. group design 2 . Participants in each group faced a scenario involving vote-buying according to the assigned treatment. The allocation of participants to each group is shown in Table 1. Note. Two candidates compete: P and Q. Candidate Q in all experimental conditions offers money. "Equal" means that the two candidates have equal qualities; "Leadership" means that P has stronger leadership qualities than Q; "Integrity" means that P has greater integrity than Q.

Materials and Procedure
We used questionnaires containing hypothetical scenarios and questions that all administered in Bahasa Indonesia. After participants indicated that they understood and agreed to the informed consent, the interviewer read the questions to the participantswhile the participants looked at the show card (Supplementary Material 2) and were shown the actual money that was hypothetically offered to the voter.
In the first part of the experiment, participants were asked to imagine that a legislative election was taking place in their area of residence, with two candidates, P, and Q, in a very tight race. Candidate P was described either as an equally strong candidate to Candidate Q, superior in leadership qualities, or integrity. The equal condition was operationalized by showing each candidate's political party and main program of legislating the law to make free education and health care for poor people (see Supplementary Material 3). The leadership quality was operationalized by giving the same information in equal conditions and additional information that Candidate P had received an award as The Best Head of Regional Legislator in Leadership in 2014. While integrity was operationalized by giving additional information that Candidate P received an award as The Best Head of Regional Legislator in Integrity from the Corruption Eradication Commission (KPK) in 2014. Afterwards, participants were asked to imagine that the interviewer was a member of Candidate Q's campaign team, offering money and instructing the participant to vote for Candidate Q (Candidate Q was the one offering monetary compensation).
Participants in the experimental condition were also assigned to one of three experimental groups with money offered. They were shown real money of varying amounts (depending on the group to which they were assigned: IDR 50 thousand, 100 thousand, or 150 thousand). Participants in the control condition did not receive the financial offer. Once participants had understood the situation, participants in the experimental condition were asked whether they would decline or accept the money offered in that circumstance. If participants were unable to provide an answer, the interviewer circled "do not know or did not answer. " Note that the question of voters' decision toward the money offered was only for participants who were offered the money. The participants in the control condition were not asked such questions. For the analysis, accepting the money was coded 1, while refusing the money was coded 0.
Following their answer regarding the money offered, participants proceeded to the second part of the experiment, in which they were given a question on candidate preference: "If you were currently in the voting booth to choose one of these two candidates, which one would you vote for?" To ensure the answer's secrecy, the interviewer gave a ballot, 2) We numbered each questionnaire from 0001 to 1220, divide them into odd (for male participants) and even (for female participants) numbers, then randomize those numbers to be assigned into 4x3 groups (4 questions of money and 3 questions of candidate quality). In each of the questionnaire, we circled the questions that should be asked for the respondents based on the random assignment and questionnaire number. Note that all of the questions representing experimental and control group were all written in the questionnaire. But enumerators only asked the circled number for each group of questions. a tool to mark the vote, and an envelope to the participants. Participants made their selection without being observed by the interviewer and then returned the filled ballot inside a sealed envelope. The amount of money, manipulation of candidates' quality, and scenario were based on a pilot study. For the analysis, the vote for Candidate Q or the vote-buyer candidate was coded 1, while a vote for P was coded 0.
Then participants were asked to complete the inhibitory self-control scale, which consisted of six items that comprise the inhibition dimension of a broader self-control scale (de Ridder et al., 2011;Tangney et al., 2004). One sample item is "I engage in activities that are actually bad for me if it is fun" (Cronbach's α = 0.56).
Because previous research has also shown that socio-demographic variables relate to vote-selling behavior, we asked participants about their socio-demographics, namely their monthly income, education (elementary and lower, junior high school, senior high school, and higher education), religion (Islam, Protestant, Catholic, Hinduism, Buddhism, other) and sex (male, female). We included these variables in the analysis to control their effect on the main variables. For all of these categorical variables, we made dummy variables with scores 1 and 0. For monthly income, we categorized ≥ IDR1.2 million as middle-high and coded 1, while income <IDR 1.2 million as lower class and coded 0. For the education variable, the category senior high school and above was categorized higher and coded 1, while 0 for education below it. For the education variable, Islam was coded 1, while other religions were coded 0. For the sex variable, we coded 1 for male and 0 for female.
Finally, an attention check was conducted. Participants were asked about how much amount of money was offered, what were the candidates that compete in the experiments, and which candidates offered the money. Only participants who could answer correctly at least two of the three questions were included in the analysis. After that, participants were debriefed and thanked for their willingness to participate in the research.
The instrument and procedure of this research had been reviewed and granted ethical clearance numbered 731/ FPsi.Komite Etik/PDP.04.00/2016 from the Ethics Commission Faculty of Psychology Universitas Indonesia.

Results
A total of 1220 participated in the study. After attention checks, 969 were found to be satisfactory (i.e. correctly answered two out of three questions) and were included in the final sample. Due to the nature of the research questions which require all groups to have received offers, the group that did not receive any monetary offer were omitted from the analysis, making the final analysis sample N = 545. The group split is shown in Table 2, while the demographic makeup of the sample is in Supplementary Material 1. All data processing was conducted using R (R Core Team, 2021) and the Tidyverse package (Wickham et al., 2019). Additionally, we used the package emmeans to probe interactions and generate the appropriate chart (Lenth, 2022). Note. N = 545 participants being offered the money. Candidate characteristic manipulation was divided into three groups: vote-buyer and competitor have equal quality (Equal); competitor has higher integrity than vote-buyer (Integrity); competitor has higher leadership than vote-buyer (Leadership).

Stage 1
To test the effect of money on voters' decisions in stage 1 (H1) and inhibitory self-control as moderator (H2), we conducted a hierarchical logistic regression with the level of financial incentive and inhibitory self-control scores as predictors of decision toward the vote-buying offer. To test H2 specifically, we included an interaction effect between the amount of money offered and inhibitory self-control. (See the full results in Table 3). We also included ethnicity (Javanese vs. non-Javanese) as control, which was found in the exploratory correlation analysis (see Supplementary Material 2) to be related significantly to participants' decisions. We also found that the amount of money offered was indeed a significant predictor of acceptance in support of H1. This effect was moderated by inhibitory self-control, lending support to H2. Note. N = 545. ISC = inhibitory self-control; OR = odds ratio. Pseudo R 2 shows greater fit after the addition of the interaction effect on all metrics.
We probed the significant interaction between inhibitory self-control and the amount of money offered in Table 4. In groups with a higher amount of money offered, the effect of inhibitory self-control towards the decision to accept the offer strengthens, with a more negative coefficient in higher offers, Rp100,000. This result suggests that stronger inhibitory self-control is necessary to resist the temptation at greater amounts of offers. For the pattern of interaction effects between the amount of money offered and ISC to offer acceptance, please see Figure 1. Rp150.000 -1.50 0.33 -2.14 -0.86

Figure 1
The

Stage 2
We tested a second hierarchical logistic regression model to investigate the predictors of participants' decision to vote for the vote-buying candidate, testing H3. We added participants' previous decision to take the money offered as well as the candidate's characteristics and the interaction between the two variables as predictors. The candidate's characteristics were contrasted with the control group, where no information was given regarding each candidate's characteristics. We added participants' gender and education as control, both of which were found to be correlated significantly with voting decisions. The results in Table 5 show that offer acceptance remains a significant predictor of whether participants will vote for the offering candidate or not. In terms of candidate's characteristics, while in the second model we found informing candidate's integrity predicted voting decision, this effect dissipated after the interaction effect between characteristics and decision to accept were introduced. Furthermore, both the interaction between offer decision and integrity as well as leadership were not significant. Thus, H3 was not supported by the data. Note. N = 545. ISC = inhibitory self-control; OR = odds ratio. Pseudo R 2 shows greater fit after the addition of the integrity and leadership, but marginal change after adding the interactions.

Mediation Analysis
Finally, we conducted a mediation analysis to assess whether the amount of money offered by candidates and its effect on the voting decision is mediated by participants' acceptance of the said offer (see Table 6). In this model, we defined the effects as the offer amount (X) predicting voting decision (Y) mediated by offer acceptance (M). All the control variables added to the previous models predicting offer acceptance and voting decisions were also added. Note. N = 545. OR = odds ratio. X = offer amount. M = offer acceptance. Y = voting decision. The pseudo R 2 indices were much greater in the full mediated effect, where acceptance was included in the model, indicating better fit.
We found that offer amount indeed has a direct effect towards both voting decision and acceptance of money offered. However, the effect of offer amount towards voting decision diminishes with the introduction of offer acceptance. This is further confirmed via Zmediation score, which is a calculation of mediation effect for categorical outcome and variables (Iacobucci, 2012), where we found the mediation effect between offer amount, offer acceptance, and voting decision to be significant (Z mediation = 2.06, p = .04)., lending support to H4 (see Figure 2).

Figure 2
The

Discussion
In vote-selling, when voters are offered money by a vote buyer, voters are asked to make an unethical decision in two stages: first, to accept or refuse the money and, second, to decide for which candidate they wish to vote. The first stage plays a vital role because the decision to accept or refuse the money could determine which candidate to vote for: the vote buyer or other candidate. And in the second stage, voters can vote for the vote buyer, but they still have a choice to vote for another candidate. Understanding these two stages furthers our understanding of how voters vote in contexts where vote-buying is prevalent. Our study has sought to contribute to studies of how voters vote in vote-buying by disaggregating these two stages in the process of voting using unethical behavior as a framework. During the first stage, our study suggests that amount of money offered could influence voters to accept the money, but this effect is moderated by inhibitory self-control. Voters who have higher inhibitory self-control tend to refuse the money offered, even when the amount is higher, and vice versa. In the second stage, our study finds that decision in the first stage influences the vote choice, but our data cannot confirm the moderating effect of competitors' quality on leadership and integrity on the relationship between the first decision and vote choice. Further mediation analysis showed that the first decision, whether to accept or refuse money, played a significant mediating role between money offered and vote choice.
The result of our study brings the need to view money as a source of temptation in a vote-buying setting, where inhibitory self-control plays an important role that may explain why voters accept or refuse the money offered. The findings parallel with studies about resisting temptation, in which people with higher self-control tend to restrain the temptation in other forms, such as food for dieting people (Hofmann et al., 2014;Myrseth et al., 2009). Either as individual differences as in our study or as situational factors resulting from self-depletion (Baumeister, Vohs, & Tice, 2007;Gino et al., 2011), self-control is a defining factor in unethical behavior. While there are some limitations, our study is among the first to study the inhibitory self-control in political behavior, seeing vote-selling as unethical behavior. Our study also further elaborates on money as a temptation, as it was coined by Hicken et al. (2014), using individual differences as variables to explain voters' decisions. Further studies may elaborate on other individual differences that relate to unethical behavior in the context of vote-buying, such as Machiavelianism, locus of control, and value orientation, as individual differences that explain voters' decisions.
For the second stage of decision, our study found no significant effect of competitor's quality vis a vis vote buyer candidate on the vote choice. This finding shows that accepting money tends to bias voters against a good quality candi date. This bias in vote choice is in-line with previous vote-buying studies which found that voters can be persuaded by money (Fionna, 2015). Note that we only have one candidate who buys votes. Further studies might examine vote choice where there is more than one candidate who offers money. Under the unethical behavior framework, the vote choice after vote-selling can also be further examined in future studies to understand the mechanism of decision for example deficiency of moral disengagement and self-regulation (Welsh et al., 2015), pluralistic ignorance (Halbesleben, Buckley, & Sauer, 2004), or other social psychological factors such as relationship and social network (Brass, Butterfield, & Skaggs, 1998) which have not been studied here. Our findings also open opportunities for further studies to elaborate on which candidate qualities-and how to present those qualities to voters-will have significant effects on vote choice. In previous studies from Liddle and Mujani (2007), leadership and integrity were found to influence vote choice, but this is not the case in vote-buying situations. More research still needs to be done to elaborate on the effect of candidates' quality on the incidence of vote-buying.
To highlight the importance of the first decision on vote choice in a vote-buying context, our research suggests that accepting or refusing money offered serves as a mediator between money offered and the vote choice. The money itself positively affects vote choice, but it is mediated by the decision to accept the money. Voters who accept the money have a higher probability to vote for the vote buyer candidate. Our findings highlight the concern about the negative effect of anti-vote buying campaigns which encourage voters to accept the money but vote for their conscience. Accepting money in the first place will pave the way to vote for a vote buyer candidate even though there is another candidate who has higher quality, and this is risking the democracy.
There are limitations in our study. The survey was intended to capture 1,200 respondents, but in the end, we can only analyze data from 545 respondents. This is because the data from a group of non-offered-money respondents are not analyzed given our focus on the respondents who are being offered the money. Even so, the data is still relatively representative of the national demographic makeup, as it is most comparable in terms of proportions of sex, religious groups, ethnicities, provinces, and age groups (see Supplementary Material 1 for a detailed comparison between our sample and national statistics [BPS, 2010]). Some observations were eliminated due to attention checks. Future studies may include the data of voters who are not being offered the money but have to find a way to identify their intention of accepting or refusing money if they were being offered the money. Another limitation is that no group received manipulation of vote buyer candidate who had high quality. For example, a candidate known for her/his integrity or leadership but offering money to buy the vote. Future studies can also consider those manipulations to answer the question of the vote choice in the election where the high-quality candidate buys votes compared to other good candidates but not buying votes.
Our findings offer insights for programs that aim to eradicate money politics in Indonesia. Anti-vote buying campaigns should inform the voters that they should be assertive about refusing financial offers in the first place and educate the voters about the negative impacts of vote-buying and selling for the public and democracy.

Conclusion
In vote-selling, we propose that voters make two decisions. First, to accept or refuse the money offered, and, second, to vote for a certain candidate. Money may affect the first decision, but the effect is moderated by inhibitory self-control. Voters' first decisions influence their vote choice. Another candidate who has higher quality in integrity and leadership compared to the vote buyer candidate still cannot strongly motivate voters who already received the money to vote for non-vote buyer candidate. Therefore, the first decision to refuse money is vital for voters to vote for a good candidate. Despite some limitations, the findings generated by this study have theoretical and practical applications for future research on vote-buying, as well as the efforts being exerted to eliminate this practice. Sekarasih, and anonymous reviewers for their generous feedback. Finally, we are grateful to Deni Irvani for helping us set up the design for the logistic regression analysis and Nick Kuipers for his proof reading.

Competing Interests:
The authors have declared that no competing interests exist.

Supplementary Materials
The Supplementary  Variable meanings and translations are explained here.