CONSISTENCY AND VOTE BUYING: INCOME, EDUCATION, AND ATTITUDES ABOUT VOTE BUYING IN INDONESIA

Abstract

This article examines ambivalence—the simultaneous holding of two or more conflicting values or beliefs about a political issue—among Indonesian citizens’ attitudes about vote buying. Using an original survey taken during the 2014 Indonesian elections, we analyse the factors related both to citizens’ normative views about vote buying and their willingness to accept gifts from candidates. A large number of citizens demonstrate ambivalence by viewing the practice as unjustified or corrupt and yet expressing willingness to accept money from candidates. We also examine the differential effects of education and income on these attitudes. Consistent with “demand side” theories of vote buying, low income creates economic pressure to accept money but does not influence normative attitudes about vote buying. Education, however, has a broader effect by influencing both normative attitudes and willingness to accept money. We consider implications of these results for Indonesian officials focused on reducing vote-buying behavior.

INTRODUCTION

The prevalence of vote buying—where candidates and intermediaries distribute commodities or money to citizens in exchange for their votes—has been known to Indonesian scholars for decades. While long viewed as a corrosive and corrupting influence on Indonesian politics, more recent scholarship has begun to focus on the nuances and complexities of vote buying. For example, Fionna (2014) found that intermediaries or “brokers” who distribute money on behalf of candidates often target the most persuadable voters and those who could be persuaded to vote in an upcoming election. ¹ Other research has examined the “demand” side of vote buying to determine which voters are most likely to prefer receiving payments from candidates as opposed to broad-based public policies (Shin 2015). Underlying much of this research is a presumption that the practice of vote buying undermines government effectiveness (Hicken and Simmons 2008) and is normatively antithetical to classic conceptions of democracy. Government agencies such as the Komisi Pemilihan Umum (or KPU) and the Kesatuan Bangsadan Politik (or Kesbangpol) spend considerable time and energy attempting to reduce the practice by distributing pamphlets to citizens and holding public events discouraging citizens from voting for candidates who offer gifts in exchange for their votes.

Less studied, however, are the attitudes regarding vote buying, and the sources of those attitudes, among the mass public in Indonesia (see Shin 2015). Research has shown that income and education are related to normative attitudes about vote buying (Halim 2013) and also acceptance of money from candidates and campaigns (Shin 2015). However, there are two distinct attitudinal dimensions to the practice of vote buying. The first is the normative dimension: whether or not the practice of candidates offering money is acceptable or justified in a democratic society. Many elites and scholars presume vote buying to be corrupt and normatively bad for democracy. This includes government officials at the KPU and Kesbangpol as well as media outlets who endeavor to shape this norm through reminders to citizens following candidate debates and other broadcasts related to Indonesian elections. The second dimension, however, involves a practical or pragmatic attitude about vote buying: regardless of their normative views, will citizens accept money from candidates if offered? It seems likely that many citizens will have consistent views across the two attitudes. For example, someone who views vote buying as ethical is likely to also accept money from candidates. However, conflict between the two values is also possible. One can easily imagine a citizen who simultaneously views the practice of vote buying as corrupt and unjustified while also feeling economic pressure to accept the money proffered by candidates and their intermediaries. Table 1 illustrates the typology created by these two separate attitudinal dimensions.

Table 1 Typology of Consistent and Conflicted Attitudes about Vote Buying

As Table 1 shows, citizens who view vote buying as acceptable can either be consistent and take the money from candidates or they can be conflicted and reject the money. Likewise, citizens who view vote buying as normatively wrong can also be consistent by refusing the money or can be conflicted by accepting money from candidates. This type of conflict or cross-pressure is common in other areas of public opinion (see, for example, Zaller and Feldman 1992; Zaller 1992; Alvarez and Brehm 1995), but little research has been conducted on the existence of conflict among Indonesian citizens about the practice of vote buying, let alone on the sources of such conflicted attitudes.

We seek to build upon the research on the micro foundations of vote buying, by exploring the prevalence of conflict between citizens’ attitudes about the moral dimension of vote buying and the practical dimension of accepting money offered by candidates. We also examine which factors best predict whether citizens are conflicted in their attitudes about vote buying. We find that two factors in particular—education and income—are strong explanations of whether voters hold consistent or conflicting attitudes, though they have somewhat different effects. Whereas past research has tended to treat education and income as measuring the same underlying concept (i.e., poverty and socioeconomic status), this paper finds that low-income voters are especially likely to report conflict between the recognition that the practice of accepting money is morally wrong and the economic reality of needing the money offered by candidates. The effect of education, after controlling for income, leads to more consistent behavior as citizens are more likely to view vote buying as morally wrong and also to reject money offered to them by candidates. Our results suggest that income and education have distinct and separate effects on attitudes about vote buying in Indonesian elections.

INCOME, EDUCATION, AND CONFLICTING ATTITUDES ABOUT VOTE BUYING IN INDONESIA

Despite being widely criticized by government officials, the media, and other elites, the practice of vote buying persists on a widespread basis in Indonesia. For example, a survey conducted by The Asia Foundation found that 34 percent of Indonesian citizens had been offered money during a recent election (Thornley 2014). The same study found that 38 percent of all citizens said they would accept money if it was offered by campaigns. Moreover, during interviews we conducted in the Spring of 2015, government officials in several large cities across Java expressed pessimism about stopping the practice in upcoming elections. Many of the programs enacted by their agencies in recent years were designed with the goal of long-term impact through socialization of younger voters (especially in Indonesian schools), with the recognition that dissuading citizens from accepting money in upcoming elections was significantly more difficult to accomplish. Thus, empirical evidence indicates that a sizable minority of Indonesians are offered money in exchange for votes, that an even larger minority is willing to accept those payments, and that a broad range of government agencies and media outlets recognize that the practice of vote buying is difficult to prevent.

The extant literature provides some insight as to why vote buying persists despite widespread public education campaigns against the practice. In particular, the literature finds that poor and less educated citizens are more likely to be offered cash or other rewards by candidates in exchange for their votes (Stokes 2005; Nichter 2008). Moreover, Stokes (2005) found that citizens with lower incomes and less education were also more likely to respond that their votes were influenced by the gifts (322). In Stokes’ theoretical framework, poorer voters value the small rewards (cash or commodities) more highly than wealthier voters, which results in parties being able to gather large numbers of votes with relatively small payments per voter. Other research has explained this relationship in terms of policy preference differences across different types of voters. According to this theoretical perspective, poor voters have a strong preference for particularistic and immediate benefits whereas wealthier voters prefer more programmatic benefits that could take longer to be delivered (Scott 1972; Desposato 2007; Shin 2013). Shin (2015), for example, asked Jakarta citizens to rank their preferences for two patronage benefits (e.g., money and jobs) as well as three more universal programs (e.g., free education). The results demonstrated that poorer and less-educated voters ranked the patronage benefits higher than jobs compared to their wealthier counterparts. Thus, one prominent explanation as to why vote buying continues in Indonesia despite the best efforts of government agencies and anti-corruption activists involves the demand for particularistic, immediate benefits from poorer and less-educated voters. This “demand side” explanation focuses on the economic pressure created by poverty as translating into a demand for money from the candidates by poorer voters.

Much of this literature views the “demand side” explanation as driven by socioeconomic status, and particularly the role of poverty in creating a demand for money or gifts from candidates. Less studied is the interaction between citizens’ economic status and their normative views about vote buying practices in Indonesia. Undoubtedly, poverty forces some voters to demand money and gifts from candidates, but are these demands for money in contradiction to their normative views about the practice of vote buying as compared to wealthier citizens? Put differently, “demand side” explanations suggest that poor voters might otherwise not accept the payment from candidates if they were not impoverished. Many citizens might want to live up to the ideals of democracy but cannot afford to do so because they need the money for themselves and their families. This would induce the conflict discussed above where citizens recognize that vote buying is normatively wrong but are also in financial situations where they feel it necessary to accept the gifts from candidates (e.g., the “conflicted accept money” category in Table 1).

Related to this question about differences in normative attitudes is the potential role of education and income on attitudes about vote buying. Many studies treat education and income as proxy variables for the same underlying concept—poverty and socioeconomic status. For example, Nurdin (2014) and Shin (2015) both rely upon education and income as proxy variables for socioeconomic status in their studies. While highly educated citizens are certainly likely to have greater economic opportunities, education and income also can have very difference consequences for citizens’ political attitudes. As Gonzalez-Ocantos, de Jonge, and Nickerson (2013) argue, the negative consequences of vote buying are not always obvious for citizens and can often be “abstract” in the sense that such consequences are not directly and immediately observed by individual citizens. Gonzalez-Ocantos and colleagues thus treat education as a proxy variable for citizens’ level of political sophistication and their awareness of these “distant” and “abstract” negative consequences (201).

Education is also likely to measure socialization into broader democratic values that recognizes vote buying as undermining democratic governance. Public opinion literature has long demonstrated that higher levels of education are associated with fundamental differences in political attitudes, including levels of political participation, acceptance of democracy, tolerance towards minority groups, and other important political beliefs (for an example, see Mujani, Liddle, and Ambardi 2011). Beyond acceptance of democracy, education is also strongly related to more qualitative understandings about the nature of democratic governance. These different meanings of democracy are often discussed in terms of “substantive” and “procedural” democratic values (see, for example, Rose Mishler, and Haerpfer 1998; Bratton and Mattes 2001; Dalton Shin, and Jou 2007; Canache 2012). “Substantive” values refer to the outputs of government and include things such as social equity, effective government, and other economic or material benefits produced by government. In contrast, “procedural” democracy involves valuing the norms and procedures of free elections as well as various freedoms and liberties for citizens in a democratic society. Citizens who conceptualize democracy in the more instrumental “substantive” terms are more likely to demand economic benefits from government, and to adjust their support for democracy in light of economic downturns, than are those who adopt the less instrumental “procedural” conceptualization of democracy.

The East Asia Barometer asked citizens about the essential characteristic of democracy and categorized responses into four categories. Two of these categories measured “substantive” characteristics such as narrowing the gap between rich and poor and providing basic necessities of life such as food, shelter, and clothing, while the other two categories measured “procedural” characteristics such as holding government accountable and the freedom to express opposition to government. Citizens with higher levels of education were more likely to choose responses indicating that democracy involved freedom to express opposition to government or citizens holding government accountable than more economic characteristics such as reducing inequality or providing basic necessities of life (data not shown). ² The East Asia Barometer dataset also showed that citizens with higher levels of education were also more interested in politics and were more likely to follow the news (data not shown). Chu, Huang, and Lu (2015) likewise demonstrated that citizens across 13 East Asian countries were more likely to conceptualize democracy in “procedural” terms if they had a college degree or higher while citizens with lower levels of education were more likely to conceptualize democracy in “substantive” terms.

Finally, beyond socioeconomic status and qualitative conceptualizations of democracy, education is also likely to increase the reception of anti-corruption messages from the government, media, and other actors that are often targeted to students. During author interviews with several government agencies and political party leaders across Central Java, East Java, and Kalimantan Island, many of the efforts at dissuading citizens from participating in vote buying in Indonesia were targeted at students in both secondary schools and college. Media commentary about the negative effects of vote buying was also prominent on major television stations and the largest newspapers in Java. Highly educated citizens are more likely to be exposed to these messages and to develop more negative views about the practice than those with less education.

Thus, consistent with past studies, we expect that both income and education will have an effect on attitudes about vote buying in Indonesia. However, these different demographic characteristics are expected to have differential effects across the normative and practical dimensions. In particular, income should have the biggest effect on reducing conflicted attitudes, shifting citizens from the “conflicted accept money” category to the “consistent reject money” category. Education, however, is expected to have a broader effect by creating more opposition to the practice of vote buying as a normative proposition as well as increasing the likelihood that a citizen rejects money when offered. This should manifest as a shift from both the “consistent accept money” and the “conflicted accept money” categories to the “consistent reject money” category.

Moreover, it is possible that there are community-based or social mechanisms that could influence citizens’ attitudes about vote buying. First, there are likely to be collective action problems where an individual citizen might believe that the practice is normatively wrong, but recognizes that most of the other people in the area are accepting the money. The voter may believe that, without coordinated action with other citizens in the area, it is impossible to stop or even reduce the practice of vote buying (Lyne 2008). Recognizing that vote buying is occurring in the area regardless of her personal views, this voter might accept the money despite her personal objection to the practice. This would certainly be more likely to occur among poorer voters, but it could also occur among wealthier voters as well. The incentive to take the money in this situation is not based on personal financial need. Rather, it is based on the idea that the money will be given (and everyone else is accepting it), therefore the person might as well get the benefit of the practice. Related to this collective action problem is the possibility that social pressure could also induce people who might otherwise view the practice as corrupt to accept the money. This would especially occur in areas where community leaders negotiated informal “contracts” with candidates and the voters in the area need to demonstrate that they fulfilled their part of the agreement. By refusing the money, the individual would signal to the community that she was not participating in the collective agreement they had made with a candidate. In fact, rejecting the money could make the person feel isolated or marginalized within society, especially if others viewed it as violating an agreement with the candidate that was made on behalf of the community.

Another mechanism that could create the perception of conflicted attitudes concerns the nature of elite discourse across Indonesia regarding vote buying. A long-standing finding in public opinion studies is that citizen opinions often follow elite discourse (see, for example, Zaller 1992; Zaller and Feldman 1992). Government agencies and candidates in some kota (cities) and kabupaten might be more unified in denouncing the practice. Indeed, government leaders in Banjarmasin and Banjarbaru (located in Kalimantan Island) were adamant about encouraging citizens both to reject money and to reject the candidates who offered it. ³ Yet other government offices, including several in Central and East Java, communicated messages that could increase the number of “conflicted” responses among citizens. Numerous government officials across Central and East Java repeated variations on the same common slogan, “take their money, but do not vote for them.” It should be noted that these messages were not intended to support vote-buying practices. Just the opposite, their goal was to reduce or eliminate vote buying by rendering it ineffective as a tactic for winning elections. Yet such messages could result in citizens who view vote buying as unethical responding that they would take the money from the candidates. This type of response is not the same kind of economic dilemma described above, but rather is a strategic consideration designed to undermine the practice of vote buying. Yet the survey responses of citizens following this strategic tactic would appear to be the same as those who are actually conflicted in their beliefs.

It should be noted that these alternative mechanisms are unlikely to be isolated to voters with lower incomes or lower levels of education. In fact, the effect of social pressure might be more common among citizens with higher levels of income who do not need the money and would otherwise reject it. These alternative mechanisms have two main potential consequences for the hypotheses described above related to income and education. First, if these collective action problems are more common among citizens with higher incomes or higher levels of education, it could attenuate the estimate of the effect of these variables in a regression model. Income, for example, might measure both the effect of the economic dilemma facing poor voters who recognize that vote buying is corrupt as well as the social pressure on those with higher levels of income who might otherwise reject the money. This would reduce the size of the income coefficient in a regression model and suggest that income has a smaller effect than it might otherwise have. Second, the effect of social pressure that is not correlated with income and education would be picked up in the error term of a regression model and also possibly be reflected in a higher constant term. In other words, even after controlling for income and education, there might still be high levels of conflicted attitudes (“conflicted accept money”) because the variables in the regression model do not adequately measure the effect of social pressure and other collective action problems. This could be the case, for example, if a sizable number of citizens (regardless of education or income) engaged in the strategic accepting of money proposed by government officials and party leaders in Central Java.

In sum, attitudes about vote buying can be multidimensional, encompassing both a normative and a practical component. People can view vote buying as corrupt or acceptable, and this viewpoint can be distinguishable from their openness to accept money from candidates. For some citizens, beliefs across these two components will be consistent. These voters will view vote buying as acceptable and will accept the money from candidates or they will view vote buying as unjustified or corrupt, and they will reject the money offered by candidates. Yet, conflicted attitudes are also a possibility with voters recognizing the corrupt nature of vote buying while still being open to accepting gifts from candidates in exchange for their votes. The analysis below examines the factors related to each component separately, estimates the prevalence of conflicted versus consistent attitudes across the two components, and finally examines the effect of income and education on conflicted/consistent attitudes about vote buying.

DATA AND METHODS

The data are drawn from a sample of households in the Banyumas and Cilacap Regencies in Central Java. The survey was administered in February 2014, prior to the national legislative and presidential elections, and uses stratified random sampling to survey approximately 20 respondents in each kecamatan in the Banyumas and Cilacap Regencies. In addition to basic demographics (income, education, gender, etc.), the survey asked citizens about preferences in both national and local elections as well as job performance among the various political parties. One obvious limitation of this study is that the two kabupatan in Central Java that were sampled in this survey obviously do not represent an archipelago of 240 million people. However, the survey was designed to capture variation on the variables of interest to this study, not necessarily to be representative of the entire country of Indonesia. In that regard, the sample does provide significant variation in attitudes about the normative and practical dimensions of vote buying in Indonesia as well as variation across levels of education and income. It is possible that mean levels of conflicted and consistent attitudes about vote buying could vary across Indonesia and future research could focus on examining and explaining this variation.

There are several approaches to measuring conflicted attitudes through survey data. First, open-ended questions allow a more direct measure of respondents’ considerations and the opportunity to assess which respondents offer conflicting considerations (see Zaller 1992, 60–62 for an example). Yet many large-scale surveys do not include these types of open-ended questions. A second approach attempts to measure conflicted attitudes by separately asking respondents about their likes and dislikes toward an issue or object (see Stoeckel 2013, for example). Respondents who simultaneously give high numbers of likes and dislikes about the issue or object are said to be more conflicted or “ambivalent.” Third, surveys often do not include either of these direct measures of conflicted attitudes, and research must rely on inferential techniques to measure ambivalence in public opinion. Statistical techniques such as heteroskedastic regression are used to measure the degree of response instability, which is used as a measure of the degree of conflicted attitudes among respondents (for an explanation of this technique, see Alvarez 1997).

Our measurement strategy uses the more direct method by asking respondents separately about the normative aspect of vote buying and the practical side of accepting the money. Embedded in the survey were two questions that asked about citizens’ attitudes about their normative view of vote buying and another that asked about their willingness to accept money from candidates. The first question approximately translates as, “according to [respondent], can giving a sum of money in order to vote for a candidate be justified?” The second question approximately translates as, “according to [respondent], if there was an offer of money to vote for the candidate, would you [Accept/Reject] the offer?” ⁴ The first question is intended to measure the respondent's normative judgment about the practice of vote buying in elections while the second measures the respondent's willingness to accept money that is offered by the candidates.

Control variables include four demographic characteristics—age, gender, income, and level of education. ⁵ Age is measured using a five-point scale while income is measuring using an eight-point scale. ⁶ Gender is a dummy variable coded 1 for male and 0 for female. Finally, education is measured as an eight-point scale ranging from no education to college or above.

Finally, the models include two dummy variables intended to measure information sources. The first dummy variable is coded 1 for respondents who received most of their political information from television and the second dummy variable is coded 0 for respondents who received most of their political information from face-to-face interactions with the campaigns. Other sources of political information were left as the baseline in the models. ⁷ Government officials in both the KPU and the Kesbang Polinmas frequently run television advertisements seeking to change citizens’ attitudes about corruption, including practices such as vote buying. Moreover, candidate debates often end with a message from the debate moderators telling citizens not to vote for candidates who offer money. For example, during the 2013 Central Java Governor election, all three of the televised candidate debates ended with the moderator telling viewers, “And remember, do not vote for candidates who offer you gifts.”

In order to examine these questions about conflicted versus consistent attitudes about vote buying, we created a typology of voters from the responses to the two questions related to vote buying (see Table 1). Respondents could provide consistent responses by saying that vote buying is both unjustified and that they would reject the money (what we term “Consistent Reject Money”) or by saying that vote buying is justified and that they would accept the money (what we term “Consistent Accept Money”). Alternatively, respondents could indicate conflicted responses by saying that vote buying is unjustified but that they would accept money (“Conflicted Accept Money”) or that vote buying is justified yet they would reject gifts from candidates (“Conflicted Reject Money”).

First, only one of the 800 respondents fell into the “Conflicted Reject Money” category. ⁸ A small majority (53.4 percent) fell into one of the two consistent categories, with a larger number being consistent in viewing vote buying as unjustified and saying they would reject the money (38.9 percent) with a smaller number saying that vote buying is justified and they would accept money (14.5 percent). The size of the “Conflicted Accept Money” category, however, was also quite large (46.5 percent) and actually represents the largest category of responses. Thus, a large number of respondents exhibited conflicted attitudes about vote buying, expressing conflict between the normative perspective that vote buying is unjustified while simultaneously admitting that they would accept money from candidates if it were offered.

To analyse which factors influence consistent and conflicted attitudes about vote buying, we created a dependent variable, coded -1 for “Consistent Accept Money” responses, 0 for “Conflicted Accept Money” responses, and 1 for “Consistent Reject Money” responses. This categorical variable was then regressed against the six independent variables described above using multinomial logistic regression. ⁹ The “Consistent Reject Money” category was left as the base outcome in the multinomial logit model. Finally, the respondents are clustered by kecamatan, which makes it likely that the errors in the multinomial regression models are not “independently and identically distributed.” Failing to account for this clustering of responses could artificially increase the statistical significance of the standard errors in the model (Primo, Jacobsmeier, and Milyo 2007). ¹⁰ We therefore estimate the multinomial logistic regression model using clustered standard errors to account for the potential problems created by the clustered nature of the data.

RESULTS

Table 2 presents the results for the multinomial logistic regression model, with the “Consistent Reject Money” category left as the base outcome in the model.

Table 2 Multinomial Logistic Regression Model for Conflicted and Consistent Accept Money Responses about Vote Buying

Note: Standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1

The coefficients in a multinomial logistic regression can be complicated to interpret, which is why we provide predicted percentages for each of the three categories below. However, each coefficient represents the effect of the independent variable on the probability that respondents will respond to that category compared to the base outcome (“Consistent Reject Money”). Thus, because the base outcome is “Consistent Reject Money” and the two columns of coefficients are the other two responses, the coefficients can be interpreted as the effect of the independent variables on “Consistent Reject Money” responses. For example, in the “Conflicted Accept Money” column the coefficient for income is negative, indicating that wealthier respondents are less likely to respond with “Conflicted Accept Money” as compared to the “Consistent Reject Money” category. Put simply, wealthier respondents demonstrated less conflicted attitudes than poor respondents for the “Conflicted Accept Money” category and the results are statistically significant at the .001 significance level. In comparison, the income coefficient is not statistically significant at conventional significance levels in the “Consistent Accept Money” column, indicating that the responses of wealthier respondents are not more likely than poorer ones to fall into this category compared to the consistent reject money category.

Contrasting the findings for income and education, the results are revealing. The coefficients for both income and education are negative and statistically significant at conventional significant levels in the “Conflicted Accept Money” column. This means that respondents with higher income and higher education levels were less likely to be conflicted in their attitudes and more likely to fall in the “Consistent Reject Money” category. In contrast, for the “Consistent Accept Money” column, the coefficients for income and education are both negative but only education is statistically significant at conventional significant levels. Thus, in this model, education appears to have a broader effect of shifting respondents from both the “Conflicted” and the “Consistent Accept Money” categories to the “Consistent Reject Money” category. Put simply, the most educated citizens are more likely to view vote buying as unjustified and to reject the money than they are to offer conflicted responses or to indicate that they view vote buying as normatively acceptable. Higher income, by comparison, appears to have a strong effect on reducing conflicted attitudes and leading citizens to reject vote buying, but does not appear in our model to significantly reduce the probability that respondents will accept vote buying.

Turning to the control variables, the coefficient for age is negative in both models and statistically significant for “Consistent Accept Money” column but not for “Conflicted Accept Money,” although the difference in the coefficients is not statistically significant at the .10 significance level (p < .386). While older respondents are not more likely to demonstrate conflicted attitudes, older citizens are more likely to offer “Consistent Reject Money” responses than they are “Consistent Accept Money” responses. The coefficients for Male were not statistically significant across the two models nor was the difference in coefficients statistically significant at conventional significance levels (p < 0.770). Thus, while age is related to two of the categories, gender is unrelated to all three of the categories of the dependent variable.

The final two control variables deal with information sources about politics and campaigns. Both variables were positive and statistically significant at conventional significant levels for the “Conflicted Accept Money” column while neither were statistically significant in the “Consistent Accept Money” column. This indicates that respondents who receive most of their campaign information from television or face-to-face communications were more likely to have conflicted responses about vote buying (compared to the “Consistent Reject Money” category), but neither were significantly different in terms of the “Consistent Accept Money” category. Thus, independent of demographics, respondents who received most of their political information from television or face-to-face communications exhibited more conflicted attitudes about vote buying than did respondents who received the majority of their information from other sources. The last column of Table 2 provides the likelihood-ratio test results for difference of coefficients for each variable between the “Consistent Accept Money” and “Consistent Reject Money” categories. The cells are the p-values for the difference in coefficients for each variable in the model. Thus, the difference in coefficients is not statistically significant at the .10 significance level for any of the variables except television.

To help contextualize and clarify these findings, the tables below provide predicted percentages for each of the three categories of the dependent variable across the values of income and education. ¹¹ Table 3, for example, presents the predicted percentages of responses for the lowest, highest, and middle categories for income. Among respondents with the least income, 55 percent demonstrated conflicted in their responses compared to 30.3 percent who gave Consistent Reject Money responses and 14.7 percent offering Consistent Accept Money responses. Among those with the highest income, however, a large majority (63.6 percent) offered Consistent Reject Money responses compared to only about a quarter who offered conflicted responses (24.2 percent), with another 12.2 percent offering Consistent Accept Money responses. Thus, the largest shift in attitudes about vote buying across income levels occurs as a shift from conflicted to consistent (reject money) responses. This strongly suggests that low levels of income, at least in part, can trigger conflicted attitudes among respondents that creates conflict between the normative perspective that vote buying is unjustified yet willingness to accept the money from candidates. This contrasts somewhat with the effect of education on conflicted and consistent attitudes about vote buying. Table 4, which presents the predicted percentages across levels of education, does show conflicted responses declining with education and also the “Consistent Reject Money” becoming the largest category response among the highest educated respondents. However, there is also a sizable decline in “Consistent Accept Money” responses as education increases (23.5 percent among least educated to 8.4 percent among the most educated). Thus, education influences responses to all three categories as compared to income, which mostly has an effect on the Conflicted Accept Money versus Consistent Reject Money categories.

Table 3 Predicted Percentages for Income

Table 4 Predicted Percentages for Education

Again, these results suggest that education has a larger effect than income on reducing acceptance of vote buying more broadly, while income has the strongest effect on reducing conflicted attitudes. The large drop in “Consistent Accept Money” responses across levels of education contrasts with the smaller effect across levels of income. It should be noted that because education and income are entered separately into the model, the results represent the independent effect of each variable on conflicted and consistent attitudes about vote buying. If education and income were simply measuring the same concept—poverty—then it is unlikely that the models would produce divergent results for different categories of responses. An analysis confirmed that the divergent results are not due to collinearity between income and education. ¹²

Finally, the combined effect of income and age on reducing conflicted attitudes about vote buying is demonstrated in Table 5. Each cell is the predicted percent of respondents giving the “Conflicted Accept Money” response by both levels of education and income. As the table shows, income appears to have the larger effect on reducing conflicted attitudes. Even among the most educated respondents, 46.6 percent of those with low levels of income still held conflicted attitudes. This high level of conflicted attitudes among respondents with higher education but lower incomes is likely due in large part to the economic dilemma described above where citizens may recognize that vote buying is wrong but feel financial pressure to accept money and other gifts that will benefit themselves and their families. In contrast, among those with the lowest levels of education, only 27.6 percent of respondents with high levels of income were predicted to have conflicted attitudes. Among those with both the lowest levels of income and education, nearly 60 percent held conflicted attitudes, compared to only about 14 percent of respondents who had both the highest levels of income and education. Thus, while education and income appear to have separate and distinct effects on attitudes about vote buying in Indonesia, in combination the two variables have substantially large effects at reducing conflicted attitudes and promoting more negative views about vote buying that lead citizens to reject money when they view the practice as normatively wrong.

Table 5 Predicted Percentages for Conflicted Accept Money by Education and Income

DISCUSSION

This paper has attempted to build on the growing literature examining citizens’ attitudes about vote buying. Using a unique survey that asks citizens their beliefs about the normative and practical components of vote buying, we examine the factors that are related to both normative attitudes and practical beliefs about accepting money from candidates. Our results suggest that education and income have separate and distinct (though obviously related) effects across these two dimensions. Income appears to have the strongest effect at reducing conflicted attitudes—recognizing that vote buying is wrong but expressing a willingness to take the money—and a weaker effect on citizens’ views that vote buying is acceptable or justified. Citizens with higher levels of income who also recognize that vote buying is wrong have more financial freedom to reject the money offered by candidates. Yet, independent of the effect of education, income does not appear to strongly reduce the percentage of citizens who view vote buying as normatively acceptable or justified.

Education, on the other hand, does appear to shape normative attitudes more strongly than income, probably through socialization processes and exposure to a greater number of anti-corruption messages from government officials and media sources. Put differently, education can lay the foundation for democratic values that change citizens’ attitudes about vote buying, but education by itself is unable to completely overcome the economic dilemma faced by many lower income citizens when they are actually offered money by candidates. This is evident in the high percentage of “conflicted” responses among even the most highly-educated citizens, especially among those with lower levels of income. Education and the socialization of traditional democratic values among Indonesian citizens is an important goal, but for many Indonesians these values are confronted by the economic reality faced by their families when offered money during an election.

These results have important implications for how governments and other anti-corruption actors attempt to address the problem of vote buying. Civic education and socialization efforts can be successful in altering citizens’ normative evaluations of vote buying practices. But they might not alter the fundamental economic dilemma facing poor voters that creates the demand for particularistic benefits from candidates. Moreover, many government officials in Indonesia are deeply familiar with this dilemma. The authors asked one KPU leader in a small Central Java city specifically about how to convince a poor person who might view vote buying as corrupt to reject the money offered by candidates. The KPU leader's response was simply, “it is impossible.” ¹³ Thus, the attitudinal conflict that arises when poverty creates economic pressure to accept money from candidates among voters who might otherwise view the practice as corrupt could represent a large impediment to policymakers trying to reduce vote buying in Indonesian elections.

This conflict would occur only among citizens who view vote buying as corrupt or unjustified. Yet, citizens are likely to vary somewhat in their normative attitudes about the role of vote buying in elections. While many of the government and media communications are uniformly critical of the practice, citizens also receive communications from others such as brokers and campaign team members that are not as negative about vote buying. ¹⁴ Nor are all citizens exposed to the government and media efforts to denounce vote buying practices, especially among poorer citizens who do not have access to television or live close to areas where government officials hold educational seminars and other public events. Thus, some citizens might view vote buying as acceptable, justified, and consistent with democratic elections. Citizens in several rural villages in Central Java, for example, explained to the authors about contracts made with candidates where community leaders in each village negotiated on behalf of the voters for specific amounts of money from the candidates. One citizen summed up the villagers’ attitudes about vote buying by saying, “the election is a time for harvest.” ¹⁵ In their view, democracy meant that politicians delivered goods and other services that they would not otherwise receive in non-election years because the elected officials forgot about (or simply ignored) their villages after the election.

Beyond the economic dilemma faced by voters with lower levels of income, our results also could reflect more strategic considerations by citizens that are promoted by government officials and even the parties themselves. As noted above, government officials in Central Java have employed an innovative tactic designed to undermine the effectiveness of vote buying by telling citizens to, “take the money but do not vote for the candidate.” Thus, at least part of what we measure as “conflicted” attitudes might simply be the effect of differences in elite messages about vote buying. These elite messages are not always unified, with some elites (such as those in Central and East Java) promoting a strategic “take the money” message while others (such as those in Kalimantan and Northern Sulawesi) promoting a more traditional message asking citizens to reject vote buying altogether. Considering the apparent incongruence between the normative attitudes of many Indonesian citizens and their willingness to accept money from candidates (e.g., “conflicted attitudes”), these elite efforts to address vote buying in Indonesia could be tailored to different citizens depending upon their personal situations. For example, the strategic “take their money” message would seem better suited for citizens with lower levels of income who may need the money offered by candidates while the more traditional (and normatively appealing) message to reject participating in vote buying would seem better suited for more affluent citizens, especially those with higher levels of education. In conclusion, attitudes about vote buying in Indonesia can be multidimensional and complex. Helping policymakers and government officials to focus their limited resources on the most effective methods for educating citizens and to otherwise reduce the practice of vote buying requires a richer and fuller understanding of the complex, nuanced nature of citizen attitudes about vote buying practices.