Introduction

Abstract

This chapter starts the background of the study, showing that an unprecedented wave of institutional changes in Indonesia has created more room for clientelistic strategies. It shows that this book aims to study the dynamics of vote buying in Indonesia’s young democracy, exploring the nature, extent, determinants, targeting, and its effectiveness. It then reviews the literature on vote buying, and particularly the debate on whether it mainly targets core or swing voters. The next section presents Indonesia’s institutional framework, explaining how it has helped shape vote buying. This section helps distinguish the Indonesian case from the conventional patterns identified in literature that largely stems from Latin American cases. The chapter then explains the research methodology used and concludes by offering an overview of the book structure.

“Cut off my finger if there is a MP in Indonesia today who got elected without doing buying votes!”

—One-time DPR member (Interview, 20 April 2014)

1.1 Background: Why Are New Democracies Vulnerable to Vote Buying?

As the third and fourth waves of democratisation have swept the world since the 1970s (Huntington, 1991), elections have truly become a global norm: more than 90% of countries in the world now elect their leaders through competitive multiparty elections (Van Ham and Lindberg, 2015; Global Commission, 2012). However, as many of the newly democratising regimes only achieved barely minimal standards of electoral competition, the early optimism about the rise of democracy has significantly waned in much of the global democratic world. A lot of the erstwhile enthusiasts soon turned into sceptics, given that electoral democratisation hasn’t automatically transformed into liberal form of democracy but beset by the curtailment of civil liberties and the feeble establishment of the rule of law (Diamond, 2002; Rose and Shin, 2001).

No less important, while almost all nations in the world currently hold multiparty elections, not all of them were able to improve the quality of its elections, that is, the degree to which elections are free and fair (Van Ham and Lindberg, 2015). If the de jure multiparty elections are plagued by a wide range of election fraud and electoral malpractice, the quality of democracy becomes problematic at best. Vote buying—the exchange of material benefits for or at least in the expectation of votes—is among electoral manipulation that has become a key component of electoral mobilisation in many young democracies (Jensen and Justesen, 2014). Ironically, instead of diminishing vote buying, the transition stages from authoritarian regimes to democracy may encourage such a distinctly non-democratic practice. In transitional democracies where democratic institutions (i.e. political parties) are still weak, the design of electoral and political institution consequently provides strong incentives for politicians to launch such strategies. These include competitive elections and multiparty system (van de Walle, 2007; Muno, 2010), electoral rules (Scheiner, 2007), decentralisation (García-Guadilla and Pérez, 2002), and credibility on decision-making process (Keefer, 2005).

Indonesia is no exception. Soon after Suharto’s authoritarian New Order regime came to an end in 1998, Indonesia’s party system entered a new, post-authoritarian era. Political parties, of which there were previously only three because of heavy government regulations, could now form freely. Consequently, a highly competitive multiparty system emerged, which coincided with the simultaneous introduction of multi-level elections from legislative to direct presidential ballots. In the wake of this unprecedented development, almost all political parties were neophytes with no political credibility (Vlaicu, 2016). Ideological divisions among political parties were also not salient. Thus, voters were often unable to differentiate political parties regarding policy positions or platforms. Meanwhile, post-Suharto’s electoral system that generated ‘candidate-centred’ elections (voters can choose candidates over parties) failed to mitigate such problems. Under these circumstances, candidates were forced to generate a personal vote and candidate-centred campaigns rather than building a party vote and party-centred campaigns. Thus, in order to stand out from competitors within their own parties, candidates were unhesitating to buy votes, as happened in numerous transitions during the third and fourth waves of democratisation.

It is increasingly clear that an unprecedented wave of institutional changes in Indonesia since 1998 has created more room for clientelistic strategies. The changes in political and electoral systems, even with the best of long-term goals to create free and fair elections as well as appropriate channels of accountability (Shair-Rosenfield, 2012: 1), crystallise one of basic questions guiding this book: what effects have political and institutional factors had on the nature of patronage politics in Indonesia? How well do institutional arrangements explain the scope, patterns, determinants, targeting mechanisms, and effectiveness of vote buying as a form of patronage distribution? This book aims to answer these questions posed by the connection shared between these twin issues of change: the changes in the political institutions after the authoritarian regime, especially in the electoral arena, and the changes in the nature and mechanics of vote buying.

However, given the burgeoning of scholarly writings on the impact of electoral systems on clientelism, the analytical focus of this book is on vote buying by arguing that particular contextual factors—especially the adoption of open-list proportional voting system as a result of the electoral reforms after the fall of Suharto’s authoritarian rule—matter greatly in explaining the ubiquity of vote buying in Indonesia. This book presents—in much more length—a wide-ranging study of the dynamics of vote buying in Indonesia’s young democracy, exploring the nature, extent, determinants, targeting, and effectiveness of this practice. Despite vote buying becoming central feature to electoral campaigns in Southeast Asian countries, particularly in Indonesia, where this practice has gained prominence in its post-Suharto’s electoral politics, most influential studies on clientelism have emerged from other world regions. This book therefore aims to fill the gap in the scholarship reference on electoral clientelism in Indonesia and to situate my findings about the country within the context of wider academic literature on the field.

1.2 Vote Buying in Indonesia’s Post-authoritarian Rule

As noted above, vote buying appears to be endemic in many recently established democracies (Jensen and Justesen, 2014; Keefer, 2005). For instance, using the 2005 Round 3 Afrobarometer survey, Andrews and Inman (2009) found massive evidence of vote buying in seven democratic countries in Africa,¹ and the Latin American Public Opinion Project (LAPOP ) released findings of the 2010 Americas Barometer surveys showing high levels of—and interesting variations in—vote-buying behaviour across the Latin American and Caribbean regions (Faughnan and Zechmeister, 2011).² Similarly, politicians in Asian countries often opt to target poor citizens with offers of money, goods, or other forms of compensation for their vote in elections. In the Philippines, for instance, vote buying has long been a major feature of the country’s elections, with an estimated 22% of its total electorate having been offered money or goods in exchange for their votes during the May 2013 elections (Pulse Asia, 2013).

One country in Asia that has attracted particular attention in terms of its vote-buying practices is Indonesia. It is difficult to find an analysis, either in the mass media or in the academic literature, of Indonesia’s current electoral politics that doesn’t mention vote buying, locally known as ‘money politics’ ( politik uang ). Despite its prominence, this issue surprisingly hasn’t received much systematic and comprehensive scholarly attention. The few examples of scholarly works on this topic are based on qualitative approaches (e.g. Choi, 2007; Hidayat, 2007; Hadiz, 2010; Aspinall and Sukmajati, 2016). Although they have significantly contributed to our understanding about vote buying, this qualitative literature is unable to measure vote buying’s scope, pattern, and effects on electoral outcomes (Gonzalez-Ocantos et al., 2012: 203). Equally, much of the existing literature on vote buying has relied on anecdotal evidence, often drawing from unproven rumours and claims (Corstange, 2012: 483). Consequently, little is known about how many voters actually sell their votes in Indonesia and whether cash handouts have discernible effects on turnout or vote choice. This book deals with these key questions that have haunted scholars of Indonesian studies for the last 15 years.³

Indonesia is a compelling case study to illuminate the dynamics of vote buying in post-authoritarian societies. This is because of its significance as the third largest democracy in the world and because it belongs to a group of Southeast Asian nations that report offers of vote buying in higher numbers than most other countries (Schaffer, 2007; Amick, 2016). Indeed, my study finds that vote buying is so widespread that it has become central to election campaigning in Indonesia (see Chap. 2). In addition to establishing the extent and effects of vote buying, this book also addresses broader questions in the comparative literature on clientelism, such as those regarding the determinants and targeting mechanisms of vote buying. Using survey data from Indonesia’s legislative elections in 2014, I examine a large number of variables generally believed to be the determinants of vote buying at the individual level (such as income, civic engagement, and political attitudes). Surprisingly, my results suggest that, among other things, partisanship (i.e. strong emotional attachment to a particular political party) is a highly significant predictor of vote buying. Party identifiers were three times more likely to be targets of such practice than non-party identifiers.

This finding is striking, given that strong party supporters should be expected to vote for their party without material incentives. But this result is only one piece of the evidence in a jigsaw puzzle I put together through this book. This finding cannot establish definitively whether voter partisanship attracts handouts or whether the reverse is true: that these benefits cause people to identify with the party that hands out cash (see also Stokes et al., 2013: 54). Despite such a potential endogeneity problem—which I discuss later in this book—my finding challenges one strand of scholarship that suggests that partisanship encourages voters to voluntarily help their parties during campaigns (e.g. Dalton, 2016; Verba and Nie, 1972). Further complicating this picture is the fact that the aggregate level of political partisanship—as expressed in levels of party identification—is comparatively low in Indonesia, constituting only 15% of my survey respondents during the 2014 election. Given these conditions of low partisanship, one major problem for candidates and parties arises: how feasible is it for candidates in Indonesia to win only by targeting partisans?

The issue of partisanship is in fact at the heart of the scholarly debate about the logic of vote buying (Dunning and Stokes, 2008; Cox, 2010; Diaz-Cayeros et al., 2012). Efforts at vote buying incur high costs, caused by both the need to establish a structure of vote brokers to deliver benefits to the voters and, of course, the requirement to raise the cash for distribution. With limited resources in hand, and in an environment in which ballot secrecy is protected, candidates are concerned with the effectiveness of vote buying in determining voting outcomes. As a result, they focus on the distribution of electoral incentives to some voters but exclude others (Stokes et al., 2013). The literature identifies two contrasting strategies in how they do so: the core-voter model and the swing-voter approach. The first posits that parties provide to core supporters in order to mobilise them to turn out on election day (Nichter, 2008; Stokes et al., 2013; Aspinall et al., 2015). The second sees vote buying as a strategy that attempts to sway uncommitted voters (Stokes, 2005). What types of voters, then, do Indonesian politicians target?

While the results of my voter surveys seem to provide evidence for the core-voter prediction in Indonesia, and my survey of politicians and brokers finds ample evidence of strong intentions among political actors to target partisan, loyalist voters, the picture is in fact complex. The data clearly show that although such voters are more likely to be targeted, in fact, overall they only make up a tiny proportion of the electorate in Indonesia. How can politicians rely on targeting only party loyalists to win in an election, given the limited number of such voters? As the election draws nearer, how do they make choices on how to spend their money once they have exhausted party loyalists? This challenge becomes even more complicated as, under open-list proportional representation (hereafter: PR) systems such as that used in Indonesia, candidates from the same party have to compete for votes between themselves to gain a seat. The small number of party loyalists is thus highly contested among co-partisans desperately seeking personal votes.

Moreover, while party loyalists are more likely to be targeted in relative terms, in absolute terms the data show that vote buying in Indonesia mostly occurs among non-partisans. If candidates and brokers express such a strong desire to target loyalists, why do they largely end up distributing so much cash and goods to so many uncommitted voters, and what sort of people get targeted for those payments? Hence, the puzzle underlying this book revolves around the question of how politicians and brokers decide which voters to target—voters who they might deem ‘loyal’ but who are in fact emotionally unattached to any party or candidate. In addition, the principal-agent problems inherent in vote buying—with agents leaking money provided by their principals—are also common in Indonesia. Regardless of such challenges, this book shows that candidates still pursue this electoral strategy with enthusiasm, with the result that as many as a third of voters across Indonesia are exposed to vote buying.

Why is vote buying so widespread despite targeting being imprecise and leakage high? In the context of the secret ballot, as in Indonesia, how can parties and candidates be sure that their investment has an effect on voting behaviour? It is fair to assume that without a traceable effect of vote buying, parties and candidates wouldn’t engage in it—either in Indonesia or anywhere else. Indeed, despite all the problems of inefficient delivery, my study finds cash handouts are surprisingly effective in producing higher turnout and vote share. In particular, I show that it is particularly the small margin candidates need under an open party list system that makes vote buying effective, its high cost notwithstanding. I show that, despite all its inefficiency, vote buying has an effectiveness ratio that is more than enough to make the difference in the tight races that occur between candidates on a single-party list. This helps explain the underlying logic behind candidates’ insistence on running vote-buying campaigns.

Overall, therefore, this book aims to explore the dynamics of vote buying in Indonesian electoral politics and how post-Suharto’s institutional arrangements explain the prevalence of vote buying. In doing so, I present systematic answers to many of the key questions that have arisen in the literature on clientelism. These questions concern the scope, patterns, determinants, targeting mechanisms, and effectiveness of vote buying. My primary research question is: what logic determines the patterns of vote buying in Indonesia? Developing an answer will require answering a set of subsidiary questions. These include, first, how prevalent is vote buying in Indonesia? Here, I aim to identify various forms and the intensity of this practice in Indonesia. Second, in order to explain the ubiquity of vote buying, I tackle questions such as what kinds of voters are most likely to ‘sell’ votes, what factors explain why some individuals are more likely than others to be targeted for vote buying, and how are they targeted? A third set of questions focus on the impacts of vote buying. In particular, I ask: how effective is vote buying in boosting greater turnout or vote share?

The rest of this introduction first reviews the literature on vote buying, and particularly the debate on whether it mainly targets core or swing voters. It then proceeds by introducing my main arguments, which the book chapters substantiate. The subsequent section presents Indonesia’s institutional framework, explaining how it has helped shape vote buying during elections. This section helps distinguish the Indonesian case from the conventional patterns identified in literature that largely stems from Latin American cases. The chapter then explains the research methodology used and concludes by offering an overview of the chapter structure.

1.3 Literature Review

1.3.1 Electoral Clientelism: Vote Buying

Following Nichter (2010), this study distinguishes electoral clientelism from the broader category of clientelism. Nichter (2010: 2) defines electoral clientelism as the distribution of material rewards to voters “exclusively during electoral campaigns.” This runs contrary to the generic, classic definition of clientelism which typically involves ongoing relationships where politicians (or indeed, other social leaders) provide assistance and benefits not only during elections (Scott, 1969; Bobonis et al., 2017; Kitschelt and Wilkinson, 2007). Hicken (2011: 290–294) lists a number of crucial aspects of clientelism, describing it as a form of relationship based on material exchange and involving contingency, hierarchy, and iteration. Muno (2010) adds two important elements, insisting that clientelism is personal and voluntary.

Accordingly, if we stick to such key features of clientelism, not all payments made during elections are part of clientelist relationships. Hicken (2011: 295) reminds us that some instances of vote buying that scholars have documented around the world actually don’t fit neatly into the classical category of clientelism. In many instances, vote buying is a one-off interaction rather than an example of an ongoing, or iterative, and mutually beneficial relationship of exchange (Kramon, 2011; Aspinall, 2014). In this study, therefore, vote buying can take the form of clientelist or non-clientelist exchanges. In order to capture both clientelist and non-clientelist forms of vote buying, I follow Schaffer and Schedler (2007) in viewing vote buying as an act which doesn’t need to involve an element of clientelism. Hence, I simply define vote buying as a last-minute effort to influence a voters’ decision in an election, typically taking place days, or even just a few hours, before a poll, by providing the voter with cash, goods, or some other material benefit.

1.3.2 Targeting Strategies

As indicated earlier, given the budgetary constraints candidates face, the question of how they determine the targets of their vote-buying strategies has become a key theme in the literature. Much scholarly theorising on vote buying involves two competing camps, that is, the core-voter versus swing-voter schools. The former holds that when they distribute cash payments or goods, parties tend to target their own core voters in order to motivate them to vote and thus increase their turnout (e.g. Nichter, 2008; Stokes et al., 2013). This form of vote buying is often referred to as ‘turnout buying.’ The second school suggests the opposite. According to this camp, parties will not waste their limited budgets on core supporters, but instead expend it on swing voters or weakly opposed voters (e.g. Lindbeck and Weibull, 1987; Stokes, 2005). This argument is based on the underlying assumption that a core voter is already committed to support the party and hence needs no further incentive to vote for it.

In this study, I examine these two dominant streams to explain how Indonesian politicians and their intermediaries distribute benefits to voters. Cox and McCubbins (1986) were among the first to outline the core-voter hypothesis. They contend that in many settings, political parties tend to allocate distributive benefits primarily to their core voters. The driving factor in the core-voter model is the assumption of risk aversion on the part of politicians. Core voters are seen as being more responsive than swing voters because politicians are in “frequent and intensive contact with them and have relatively precise and accurate ideas about how they will react” (Cox and McCubbins, 1986: 379). In their model, risk-averse politicians are unlikely to target swing voters and supporters of their opponents because these two groups of the electorate are riskier bets. Cox and McCubbins coined the term maintenance buying to reflect the fact that channelling benefits to core supporters is a rational strategy for a party seeking to maintain a long-term relationship with supporters. Gans-Morse et al. (2014: 4), on the other hand, called it a rewarding loyalist strategy to lock in core voters who might otherwise defect.

Focusing on the individual level, Nichter’s (2008) study developed a strong rationale for Cox and McCubbins’ argument, but with a slightly different focus and argument. While Cox and McCubbins (1986) emphasised the role of risk aversion in this strategy, Nichter (2008) argued that passive supporters are substantially more likely to receive electoral incentives due to the primary goal of vote buying being voter mobilisation (as opposed to persuasion). In his model, vote-maximising politicians don’t try to change voters’ preferences, but—and he uses the term turnout buying—their goal is to increase turnout among their supporters, some of whom may not be sufficiently militant to go to the ballot box at all costs. In addition, Nichter (2008) argues that turnout buying is much easier to monitor for parties and their intermediaries.

A study by Diaz-Cayeros and his colleagues (2012) also found empirical support for the core-voter hypothesis. In their work, however, the focus is on the endogeneity of partisan loyalties to material benefits. They argue that parties, especially in Mexico where their study was developed, tend to target loyal supporters to maintain their electoral coalitions over time. They argue that party machines still find it in their interest to target party loyalists, particularly in the presence of fear that if ignored the loyalists might defect. Finally, Stokes and her collaborators (2013) brought a different emphasis to the debate by arguing that this party loyalist strategy is used neither for systematically buying votes nor for purchasing turnout, but it is simply a manifestation of rent-seeking behaviour by electoral brokers. They argue that broker predation is the driving force behind the tendency of targeting party loyalists, as this strategy allows brokers both to get a higher profit margin from the funds given to them and to consolidate their position in their own patronage networks.

In contrast to such views, Lindbeck and Weibull (1987) proposed the swing-voter thesis. The main proponents of this strand suggest that distributive benefits will be more likely to be targeted at swing voters because it is these voters who determine the outcome of an election (Lindbeck and Weibull, 1987; Dixit and Londregan, 1996; Stokes, 2005). Parties need to address the short-term interests of voters outside of their ideological and social core constituency to expand their base at election time (Diaz-Cayeros et al., 2012). Core voters, by contrast, remain supportive of their party even if material benefits are cut off (Stokes, 2005). Therefore, the rationale behind the swing-voter strategy is simple: swing voters’ electoral decisions might be affected by gifts, while core voters are unlikely to be. In the swing-voter logic, to reward loyalists who are close to the party or candidate is to waste limited resources. In the same vein, the swing-voter hypothesis also predicts that politicians will not reward opposition supporters who are too ideologically distant from them to be persuaded by gifts to change their electoral choice.

In some of her earlier work, Stokes (2005)—in drawing from Dixit and Londregan (1996)—argued that parties avoid investing in core voters because the latter cannot credibly threaten to defect from the party. “Such a threat would lack credibility: the party knows that the loyal voter, even without rewards, is better off cooperating forever than defecting forever” (2005: 320). In her 2005 model, party machines predominantly favour swing voters, or even those swing voters who are slightly opposed to their party because only these swing voters can credibly threaten to vote with their conscience if they aren’t swayed by the offer of material inducement. Alternatively, a typical element of the swing-voter hypothesis is the logic of vote maximisation, in which favouring swing voters over core voters increases the prospect of electoral victory. As Diaz-Cayeros and his colleagues (2012: 3) put it, for the swing-voter hypothesis, “swing voters are often equated with the closeness or margin of the victory.” This is largely because in order to win elections, parties cannot exclusively rely on their loyalist voters, but they also need to persuade swing voters who are indifferent to the rival parties. Hence, investing a large amount in gifts to swing voters, in this view, can be decisive in determining electoral outcomes.

So how does Indonesia fit into this debate between proponents of the core-voter and swing-voter models of vote buying? Or does Indonesia take a different path altogether? It is worth noting that despite stark differences between the swing-voter and core-voter arguments, both camps typically assume that it is the party that is doing the vote buying (Aspinall et al., 2015 , 2017; Kramon, 2013). The problem with this sort of analysis in the context of Indonesia’s democratic transition is that, although political parties still count in legislative elections, it isn’t parties but candidates with networks of informal brokers who play the key role in organising grassroots electioneering and, therefore, vote buying. Candidates not only campaign for their party but also against candidates of the same party. As I will show below, this circumstance has significant implications for the patterns of vote buying and the targeting mechanisms used.

In addition, many scholars of vote buying have assumed that the key parameter parties use to identify recipients when distributing benefits is voters’ ideological or partisan proximity to the machine or to its opponents (Dunning and Stokes, 2008: 3). The Indonesia context, however, makes this assumption problematic. As noted above, the number of party loyalists is comparatively low, and party organisation isn’t well organised. Many of the parties are also not clearly ideologically differentiated from their rivals. Thus, the Indonesian case displays significant differences from the context in which much of the literature on vote buying was developed. This book, therefore, attempts to go beyond just testing the two dominant positions—and beyond locating Indonesia in terms of a ‘choice’ in the swing versus loyalist targeting debate. It aims to do justice to Indonesia’s complex electoral dynamics and—in turn—use the findings to inform the comparative debate on vote buying.

1.4 The Argument in Brief

This book argues that Indonesia’s open-list proportional voting system—as a result of the post-Suharto’s electoral reforms—plays a crucial role in explaining the ubiquity of vote buying. Under this system, candidates must compete against their party peers for personal votes, and they only need to win a small slice of the votes to defeat their internal rivals. This means that despite vote buying’s inefficiency and small effects, in close contests such as in Indonesia, vote buying can still make a difference to a candidate’s odds of victory. Even if offers of money influence the vote choice of ‘only’ 10% of voters, this figure is high enough for many candidates to clinch a win.

This book demonstrates the centrality of vote buying to election campaigns in contemporary Indonesia. It shows that such practices aren’t only prominent in national legislative elections but in local executive elections as well. Based on survey responses on all measures of vote buying, as will be thoroughly discussed in Chap. 2, the estimated proportion of people engaging in this activity lies between 25% and 33% of voters. These figures, however, define a range, rather than an accurate point-estimate, of vote buying incidents in Indonesian electoral politics. In the legislative election in 2014, there were around 187 million registered domestic voters. Hence, the range of between 25% and 33% would mean an estimated 47 million to 62 million voters nationwide were offered material benefits in return for their vote. If we rely on the highest estimate, one out of three voting-age Indonesians was personally exposed to vote buying, making Indonesia the site of the third largest reported frequency of vote buying in the world, as measured in recent surveys. High levels of patronage distribution are also pervasive in local executive contests. My local elections dataset measures vote buying in terms of its acceptability among voters rather than its frequency; we can also use this measure as a proxy for those likely targeted by the practice. Utilising a rich vein of voter data from 2006 to 2015, I show that the acceptance level of vote buying is comparatively high, with four out of ten Indonesians finding it acceptable for politicians or their brokers to distribute cash or gifts as part of campaigning in local elections.

Given that vote buying is so widespread in Indonesia, it is crucial to identify the determinants of the practice. Specifically, whom do candidates target with their vote-buying efforts? I show that the consistent findings of multivariate analysis based on pre- and post-legislative election nationwide surveys suggest that party-based partisanship (or party identification) is among the strongest predictors of vote buying. Put differently, my study found that the closer the ties of a voter to a political party, the more likely he/she was to receive offers of vote buying. On the surface, the underlying rationale seems to be simple, as suggested by many proponents of the core-voter school: by targeting party loyalists, candidates and brokers reduce the risks of vote buying; such voters are the most reliable and have the greatest electoral potential. Clearly, the results from my individual data are in line with expectations from the literature on the core-voter model, suggesting that party loyalists are an attractive target of electoral clientelism in Indonesia.

But a closer look at the data in the framework of the specific Indonesian context raises more complex issues and questions. First, in the context of an open-list proportional representation system, where voters have at least some influence on the order in which a candidate within the same party is elected, such as that in Indonesia, partisan voters are highly contested among co-partisans (i.e. candidates of the same party). In a voting environment where securing seats doesn’t so much depend on defeating candidates from different political parties but on winning against internal party competitors (Selb and Lutz, 2015: 335), the candidates need to translate a voter’s partisanship—their support for their party—into a personal vote. The link to clientelist strategies is clear: in order to outdo their fellow intraparty candidates, candidates need to differentiate themselves and one way to do so is by buying votes (Aspinall and Sukmajati, 2016: 13). Thus, as the main actors of vote buying, candidates focus on party loyalists because they are the real battleground in open-list systems—adding an important nuance to typical core-voter arguments such as risk aversion and turnout mobilisation.

Second, if it is true that party loyalists are more likely to be targeted for vote buying, how feasible is it for candidates to win only by targeting such partisans? It is important to note that the number of partisan voters in Indonesia is comparatively small. Only 15% of my national survey respondents admitted being close to any political party during the run-up to the 2014 election. Equally, the number of voters who voted for the same party in the 2014 and 2009 elections was only 22% of the total electorate. Conversely, the number of non-partisans, defined as those who don’t feel close to any political party, is extremely large by any standard (85%). Given such limited mass partisanship, candidates can quickly exhaust the supply of voters if they decide to target only party loyalists with their vote-buying efforts. Indeed, as noted above, my voter surveys showed that the vast majority of vote buying—in absolute terms—happens among uncommitted voters. Thus, despite actors’ insistence that they are targeting partisan voters, the reality is that they are mostly paying out benefits to non-partisans. This outcome appears to be a flow-on effect of the small number of voters with close emotional links to parties.

In its focus on understanding candidates’ strategy of selecting targets of vote buying, this study offers an additional explanation to the scholarly debate between core- versus swing-voter models by combining an emphasis on the core-voter strategy with an emphasis on personal networks, such as are widely used by candidates in Indonesia. By doing so, this study captures the gap between the declared efforts and intent of candidates and brokers to target partisan voters, and the reality that most benefits are distributed to voters who don’t in fact feel close to any party but who are instead embedded in personal clientelistic networks which are linked, often through long chains of personal connections, to the candidate. These networks, which include but often vastly exceed the constituency of deeply committed party loyalists, function as the primary target area of brokers and candidates, explaining how both core and swing voters receive benefits. This explanation contrasts with the assumptions that typically underlie both the swing- and core-voter models. In these models, particularistic rewards are distributed in a highly targeted way to specific types of voters guided by the partisan preferences of the recipients to the machine or its opponents (Dunning and Stokes, 2008: 3). In my analysis, personal connections are key. While my argument differs from the dominant literature on vote buying, it complements earlier works on the significance of personal networks in facilitating clientelistic practices.⁴

This study calls the strategy used by Indonesian candidates a ‘personal loyalist’ strategy, insofar that it targets persons not on the basis of their partisan affiliations but as identified through personal networks. Though candidates using this approach will still target partisan voters (who will typically be connected to them through party or other personal networks), such voters have been personalised in the sense that what counts for the candidate isn’t only their loyalty to the party but also their loyalty to the individual candidate within the party. The reliance on personal networks rather than party loyalties and linkages can be expected to be most prevalent in settings where political parties are largely absent in election campaigns and where partisan ties are weak. In Indonesia, the adoption of an open-list PR system in the beginning of its transition to democracy played a significant role in encouraging the development of such a context. The open-list system shapes candidates’ strategies in three ways: (1) they are forced to compete against internal competitors for personal votes; (2) they must rely on personal networks rather than the party structure; and (3) they only need to win a small slice of the voters to defeat their co-partisan rivals. With limited resources in hand and dealing with large constituencies, they are more likely to invest in areas that have traditionally been viewed as their party’s strongholds—which they think would provide the largest pool for their personal votes too. But given that party constituents are limited and highly contested among co-partisans, every candidate seeks to personalise their party constituents in the attempt to get the most intraparty votes. Although voters might have a sense of loyalty to a party, under open-list PR systems, they still can vote for different candidates within that party, meaning that their personal choice of a candidate is highly consequential for determining which candidate wins. Even voters who simply vote for a party without indicating any preference for an individual candidate, while helping to boost the chances of that party gaining a seat, will have no direct impact on determining an individual candidate’s personal prospects of victory, since the open-list system requires parties to allocate the seat only to the candidate who receives the most personal votes. Accordingly, candidates define their so-called base voters not only on the basis of past voting record but also on the basis of personal connections. Such personal connections typically include a candidate’s birthplace, kinship, ethnic and religious networks, or even simply in terms of receipt of past patronage. Moreover, candidates tap into informal brokerage networks. In short, where personalised electoral systems focus the competition on intraparty contests, candidates try to personalise their party’s captive voters, prioritising personal connections mediated by brokers.

In its implementation, however, this personal loyalist strategy runs into various difficulties. First, most candidates and brokers tend to overestimate the number of partisan, loyalist voters. This is in part because they view past voting patterns for parties as a predictor of partisan voting behaviour in the current campaign. The latter works in some cases (i.e. some parties do have clearly defined strongholds), but overall there are strong fluctuations in Indonesians’ voting behaviour. Second, loyalty is an amorphous concept in the Indonesian context. It has multiple dimensions which include partisan terms but also include kinship, religion, and ethnic ties as well as patronage loyalties. Accordingly, candidates and brokers typically misidentify non-partisans as loyalist supporters because they misinterpret personal connections as partisan leanings. This confusion over which ‘loyal’ voters to pursue makes brokers dispense benefits to swing voters they falsely believe to be core voters. Third, brokers have strong incentives to shirk due to principal-agent problems between candidates and brokers (Aspinall, 2014). Many candidates pour cash handouts en masse but invest little effort in monitoring and disciplining brokers. Lastly, in addition to the brokers’ rent-seeking behaviour, the problem of targeting could partly be a story about agency loss between voters and brokers, and between voters and politicians. Many of the people who are selected through personal networks are in fact not even loyal to the candidate.

But if vote buying is often so misdirected and susceptible to broker predation, why do candidates invest so heavily in it? Recall that vote buying is ubiquitous in Indonesia. If such clientelist exchange is truly inefficient, how can it have an impact on electoral outcomes? In the comparative literature on vote buying, measuring the effect of vote buying on voting behaviour has two main dimensions: (1) it is assessed whether cash handouts are effective at producing higher turnout and (2) it is measured whether they have an impact on determining voting choice. Despite the misdirected targeting and the unreliability of brokers, my study demonstrates that vote buying has a significant effect on voter turnout. My voter survey reveals that respondents who experienced a vote-buying attempt were more likely to vote (81%) than those who didn’t (74%), and this difference is statistically significant at conventional levels (p = 0.017).

Similarly, my study shows that the estimated effect of cash handouts on vote choice is up to 10%, meaning that 10% of voters cast their vote as a direct response to a gift of cash or goods. In this seemingly low number, however, lies the key to vote buying’s attractiveness. As elaborated in more detail below, in a highly competitive open-list system like Indonesia’s, where candidates only need small margins to beat co-partisans, that 10% can be a deciding factor in an election. Utilising official election statistics to assess the competitiveness of the 2014 legislative elections, the average winning margin for candidates when defeating party rivals was only 1.65%. Therefore, many candidates enthusiastically pursued vote buying because such a strategy could be a potential game changer. My dataset of electoral district surveys also shows that as electoral races grew tighter (marked by smaller margins of victory), the more likely a voter was exposed to vote buying, and vice versa.

Note that chasing a slim margin of victory isn’t the only explanation for why vote buying is so widespread, given there are many competitive elections in many countries where the vote margins are small, and still politicians don’t buy votes. There are two more plausible interpretations for why politicians might still prefer vote buying, even if such investment only produces low returns, and even if they aren’t in a particularly close race. The first is some sort of prisoner’s dilemma-type coordination failure (Chap. 7). Candidates would stand to gain more if they don’t engage in vote buying. But the consequences and risk that if they don’t and others do, mean that they will lose the game. While participating in vote trading may be inefficient, and may not result in victory, not buying votes is a guaranteed losing strategy (Hicken et al., 2018). Second, judged by its objective effectiveness in mobilising votes, vote buying isn’t a quite reliable strategy although it relatively still is more efficient than all the other feasible alternatives. As will be discussed in Chap. 7, the narrative among candidates and brokers is that vote buying is perceived to be more significant in driving votes, relative to other mobilisation tactics.

This finding further strengthens the conclusion that vote-buying patterns in Indonesia differ from both the core- and swing-voter models. Core voters, in the sense of party supporters, are indeed primary targets of vote buying but, in total numbers, uncommitted voters receive most of the benefits, whether intentionally or not. This is largely because of the institutional context that shapes the dynamics of vote buying in Indonesia. The conventional literature on the swing- versus core-targeting model is framed by a context that is quite different from that in Indonesia. The difference is particularly evident in the electoral system (Indonesia has an extreme version of an open party list system) and, correspondingly, the degree of party identification (which is extraordinarily low in Indonesia). This drift towards a more candidate-centred electoral system has personalised voting and vote buying and has undermined parties’ role in elections and eroded party loyalty among voters. Given that the high levels of vote buying in Indonesia are closely linked to the institutional setting that produced them, I suggest that vote buying will continue to be pervasive as long as the existing electoral framework and related socio-political settings persist. Instead of diminishing incentive for vote buying, in an environment where parties are weak and voters determine the fate of individual candidates, the adoption of an open-list voting system in Indonesia is creating a situation where vote buying has become rampant, and such practice then leads to serious failures of representation and accountability in the country.

1.5 The Institutional Framework: Party System and Electoral Rules

As noted above, candidates’ decisions to launch vote buying need to be examined and explained in relation to the institutional and structural context within which they operate. This section highlights such institutional frameworks, notably as they relate to the party system and the open-list electoral system. Both have tended to encourage vote buying.

1.5.1 Political Parties and Party System

As indicated above, Indonesia’s party system has entered a new chapter after the resignation of Suharto in 1998. Post-Suharto period has also experienced simultaneous multi-level elections from legislative elections to direct presidential elections. All legislative elections ranging from those for the national legislature known as the People’s Representative Council (Dewan Perwakilan Rakyat, DPR) and the Regional People’s Representative Councils (Dewan Perwakilan Rakyat Daerah, DPRD)—both at the provincial and municipal/district levels—and the upper (but in practice only advisory) chamber known as the Regional Representative Council (Dewan Perwakilan Daerah, DPD) occur simultaneously. Up to and including 2014, they preceded the presidential election by approximately three months (Allen, 2015).⁵ All legislative elections at different levels are held in multimember constituencies divided into multiple electoral districts known as daerah pemilihan (electoral districts). In 2014, there were 77 national districts that varied in size between 3 and 10 seats (the electoral system will be explained in more detail in the next section).

Early in the post-Suharto period, there was an explosive growth of parties. The political elite has since then tried to reduce the number of political parties over time through various registration requirements, which have been gradually tightened. In 2014, only 12 national parties were allowed to compete in the national legislative election (down from 38 in 2009, 24 in 2004, and 48 in 1999). Further, the elite has closed the door to independent candidacies in legislative polls—persons who aren’t nominated by these nationally registered parties aren’t allowed to run.⁶ These limitations notwithstanding, the number of national parliamentary parties remains significant by international standards and contributes to the competitiveness of the party system.

Political parties in the post-authoritarian Indonesia are diverse and can be classified into various categories in terms of their religio-ideological orientation and political purpose. Perhaps the oldest binary category developed by Indonesian scholars is that of secular parties on the one hand and Islamic parties on the other (Liddle and Mujani, 2010). Among the 12 national parties in 2014, 5 parties can be viewed as Islamic, while the rest can be categorised as ‘secular.’⁷ Within these categories, parties aren’t homogenous, and there is a range of ideology and policy platforms. Islamic parties, for instance, can be defined as those that either explicitly claim Islam as their party ideology, or which don’t do so but still draw most of their support from long-established Islamic organisations. Secular parties, likewise, have a range of historic and cultural differences.

Alternatively, but relatedly, Indonesian parties can be divided by their being part of a socio-cultural cleavage (‘aliran’), or by their catch-all orientation (Aspinall and Sukmajati, 2016: 16–17; Mietzner, 2013). The former includes those which had their roots in the Islamic community or, by contrast, in historically developed nationalist groups. For example, Partai Amanat National (PAN , National Mandate Party) benefits from its close links to the largest modernist Muslim organisation, Muhammadiyah, and its once solid base among religious-minded urban middle classes. Partai Kebangkitan Bangsa (PKB ; National Awakening Party) is associated with the biggest Islamic organisation, Nahdlatul Ulama (NU), and draws on support especially among underprivileged traditionalist Islamic communities in rural Java. Despite profiting from close ties and associations with Islamic organisations, these two parties present themselves as pluralist. At the more conservative end of the spectrum, Partai Keadilan Sejahtera (PKS , Prosperous Justice Party) came out of the Tarbiyah, Muslim Brotherhood-inspired campus movement and draws its strongest support in major urban centres. Partai Persatuan Pembangunan (PPP , United Development Party) was formed as a result of Suharto’s fusion of Islamic political parties in 1973 and has maintained an Islamist stance on important policy issues. On the other hand, Partai Demokrasi Indonesia Perjuangan (PDI-P , Indonesian Democratic Party of Struggle) describes itself as ‘nationalist,’ promoting pluralism and protection for minority groups. It therefore has particularly found acceptance among abangan , socio-economically lower-class nominal Muslims, and in areas with predominantly non-Muslim populations.

Contrary to the aliran -based parties, catch-all parties seek to maximise votes by attracting “as wide a variety of social interests as possible” (Gunther and Diamond, 2001: 26). This leads the catch-all parties not to appeal to any particular social group or constituency. The most-cited example is Golkar, the political machine of the Suharto regime. It claims to serve the interests of the entire nation and styles itself as “a non-aliran, non-sectarian and non-ideological party” (Hatta, 2000 quoted by Tomsa, 2008: 96). Furthermore, Aspinall and Sukmajati (2016: 17) mention the important and growing subcategory of ‘presidentialist parties,’ in which parties only serve as a political machine for their founders seeking presidential office (Samuels and Shugart, 2010; Ufen, 2006). The examples include Susilo Bambang Yudhoyono’s Democratic Party, Prabowo Subianto’s Gerindra (Gerakan Indonesia Raya, Great Indonesia Movement Party), Wiranto’s Hanura (Hati Nurani Rakyat, People’s Conscience Party), and Surya Paloh’s NasDem (Nasional Demokrat, National Democrat) Party. The presidentialist parties rely heavily on personalist appeals of their main figures—but this is often coupled with patronage delivery and populist policies to attract voters.

While presidentialist parties have often been described as lacking ideological commitment, this absence isn’t a monopoly of the presidentialist parties. Ideological divisions among political parties are generally not strong—with the exception of the schism between those who want a stronger role for Islam in state organisation and those who don’t.⁸ On most other issues, party positions are near-arbitrary, and shaped by vested rather than ideological interests. Thus, voters are often unable to differentiate political parties regarding policy positions or platforms. Given that parties tend not to compete on programmatic grounds, parties and office-seeking politicians are viewed primarily as personal distributors of private rather than public goods (Mueller, 2011). Confronted by these challenges, almost all politicians (at least those running for legislative seats) I encountered tried to focus voters’ attention on personalities, instead of party, as a voting cue. To be fair, some parties or candidates attempted to go through the motions of presenting programmatic promises in the 2014 elections, but their pledges commonly lacked credibility (Keefer, 2007). Instead, they typically quickly resorted to various clientelistic strategies to appeal to voters.

Another institutional factor that characterises political parties in Indonesia and contributes to the widespread practice of patronage politics is the weakening of the parties’ roots in society as evidenced in the general decline of party loyalty. As will be discussed more comprehensively in Chap. 5, party affiliation has decreased significantly over the past 15 years from about 86% of voters who felt close with any party in 1999 to only 15% in 2014—a low figure by international standards. These statistics correspond with the declining trend of party membership in Indonesia from 2004 to 2014 recorded by two polling institutes I am affiliated with, Indonesia Survey Institute (LSI ) and Indikator Politik Indonesia. As shown in Fig. 1.1, the trend reports just how steep the decline has been in Indonesia from around 10% in August 2004 to roughly 1.5% in mid-2014.⁹

Fig. 1.1

Trend of party membership in Indonesia, 2004–2014 (%). Sources: A series of surveys from April 2004 to September 2012 belong to LSI; while surveys in December 2013, August 2014, and October 2014 owned by Saiful Mujani Research and Consulting (SMRC); Surveys in January 2014, February 2014, April 2014, May 2014, and June 2014 by Indikator

Of course, the decline of party membership isn’t unique to Indonesia, and nor is the current level of party membership extraordinarily small by international comparison (Mietzner, 2013: 44–45). But the decline in both party identification and membership reinforces the notion that Indonesian elections are getting more candidate-centred, and provides another incentive for candidates to engage in private clientelist exchanges.

Some scholars maintain that the sharp decline of party identification and party membership correlates with electoral volatility, which denotes the extent of voters’ inclination to switch their support between elections (Mujani et al., 2012; Mietzner, 2013; Tomsa, 2014). This electoral volatility, while not high compared to other new democracies, is significant. In 1999, the top five parties accounted for more than 80% of the vote, and PDI-P won the election gaining 33.74%. In 2004, the share of the five biggest parties dropped sharply to just 66%, and Golkar came out as the champion with ‘only’ 21.58%. The declining trend continued in 2009 with the top five parties at just 61%, and the winner Democrats received only 20.85%. In 2014, the share of the main parties increased somewhat to 62%, but the victorious PDI-P got only 18.95%. Thus, the socio-political and institutional settings of the party system—with its increased focus on catch-all appeals as well as highly competitive interactions between no less than a dozen parties and thousands of their candidates—have fuelled increasing personalisation and loosening ties between parties and voters.

1.5.2 Electoral Rules and Its Implications

Indonesia’s electoral institutions have also affected the extent of patronage politics, especially in terms of candidates’ choice of strategy. It is well established in the comparative literature that electoral system design can have a large impact on candidate strategies. As Hicken explains (2007a: 49), “all else being equal, where electoral systems limit voters to a single choice among parties, as in closed-list proportional representation systems, candidates are more likely to rely on party-centred strategies.” And indeed, when, in 1999, Indonesia adopted a fully closed-list system, competition took place primarily between parties. Voters cast a ballot for a fixed list, with the candidate ranking determined by the party. Candidates were therefore predominantly concerned with their positions on party lists because those positions would determine their electoral prospects. A universally recognised term to illustrate the significance of candidates’ list positions was nomor topi (lit. ‘hat number’), describing those who occupied high positions on the list and thus had a higher chance of winning. In contrast, lower-ranked candidates were called nomor sepatu, (lit. ‘shoe number’), denoting rankings at the bottom of the list. There were numerous reports of wealthy candidates purchasing winnable slots on party lists by bribing party leaders.¹⁰

Partly in response to these internal bribery dynamics, Indonesia in 2004 applied a semi-open proportional system, enshrined in Election Law No. 12/2003. Although voters were allowed for the first time to express their preference for a particular candidate, the law still allowed for heavy party control of candidates (Sherlock, 2009). In order to get elected outside their order of the party list, lower-ranked candidates had to receive an individual vote equal to or above the full party quota required to secure a seat in their respective electoral district (Allen, 2015). If they didn’t achieve this, the seat would go to the candidate placed highest on the party list (Sherlock, 2009: 6). Given the difficulty of meeting this requirement, only 2 out of 550 members of parliament were elected by achieving an individual vote which reached the quota; the rest entered the legislature via the party list.

As shown in Table 1.1, under the 2004 election law, party votes and highly ranked slots on the lists counted a great deal in determining which candidates were elected. Again, this is largely because most candidates were unable to achieve the individual quota, handing the seat to those who attained highest positions on their party lists (Sherlock, 2009). Regardless of this limitation, Allen (2015: 76) called 2004 an ‘important moment of transition’ since for the first time Indonesia introduced optional preference voting. Despite the difficulties candidates had in achieving the requirement of a full quota, the semi-open PR system began to drive candidates to get elected on the basis of personal vote.

Table 1.1 Party list position of elected national parliamentarians

In reaction to the low numbers of candidates being elected in 2004 on a full quota, the law was changed to reduce the quota for the 2009 elections. Instead of having to obtain a full seat quota, candidates only had to achieve 30% of it to secure a seat independent of the party list (Butt, 2016: 8; Sherlock, 2009: 6). However, the Constitutional Court annulled the rule and introduced a fully open-list proportional system. According to the Court’s verdict, seats won by a party had to be handed to that party’s candidates who obtained the most individual votes. This new electoral system, introduced by the Constitutional Court, had strong repercussions for candidates and their strategies. In general, fully open-list PR systems provide a strong motivation for politicians to build personal appeals and networks since their victory (Hicken, 2007a; Allen, 2015). Indonesia was no exception in this regard. Since 2009, many candidates have campaigned for personal votes without relying heavily on their positions on the party list, and an increasing proportion of lower-placed candidates have succeeded in being elected (see Table 1.1).

Given the short time between the Court’s decision and the 2009 elections, however, candidates at that time didn’t have enough time to fully switch to a personality-centred campaign strategy. By contrast, in the 2014 elections—which also applied a fully open-list PR system—they had sufficient time to adjust to the system and prepare their strategies (Aspinall and Sukmajati, 2016: 13). In this context, it is important to note two patterns emerging under the fully open party list regime: first, despite the open-list system making elections more candidate-centric, all candidates are still concerned with party votes in their constituencies. As Aspinall argued (2014: 549), “the number of seats that each party wins in a district is in proportion to the combined votes for the party and all its individual candidates there.” Accordingly, each candidate has an interest in enhancing (or at least stabilising) the party’s overall vote and thus the number of expected seats. Doing so increases his or her prospect of winning one of those seats (Samuels, 1999: 495). Second, it is generally rare for each party to win more than two seats in any given electoral district. With many candidates believing—rightly or wrongly—that they can forecast the number of seats their party will win in a specific area, the focus of competition moves from an interparty contest to rivalry between candidates of one party over its expected number of seats (Richard Sualang, interview, 26 April 2014). As a result, the pressure to collect personal votes among co-partisans increases the incentives for individual candidates to differentiate themselves from rivals on their own party list, including by buying votes and establishing a personal campaign team.

Preference ballots generally increase the degree to which candidates are elected on the basis of individual votes (Carey and Shugart, 1995: 417). Indonesia has become a particularly prominent example of this trend. To gauge the extent to which Indonesian elections are candidate-centric, I compare total party votes and candidate votes between the 2004, 2009, and 2014 elections. Although in 2004 the electoral system had a more restricted open-list system in 2004 compared to 2009 and 2014, voters equally had the option of indicating a preference for individual candidates in all of these last three elections (Sherlock, 2004, 2009; Butt, 2016).¹¹ The Electoral Commission (Komisi Pemilihan Umum, KPU), the body that organises elections in Indonesia, itself had no official results for the total share of votes collected by all candidates across political parties competing in 2004,¹² but Kevin Evans (2004) manually collected the data by comparing the share of votes for both the party and candidates relative to those who voted for parties only. On average, 46% of voters cast their votes by marking both the party and a candidate, though the proportion of personal votes varied across the 24 parties competing in 2004.¹³ Similarly, a nationwide survey conducted by the International Republican Institute (May–June 2008) estimated that of those respondents who were aware that in 2004, besides voting for a party, a voter could also vote for an individual candidate, 47.2% reported voting for party and candidate from the same party, a total of 35.4% said they voted for the party only, and 17.4% had forgotten what they did. A survey organised by the International Foundation for Electoral System (IFES) provides a slightly higher estimate: 52% of Indonesians during the 2004 legislative elections indicated a preference vote for individual candidates (Wall, 2004 quoted by Sherlock, 2009: 8).

In 2009, when a fully open-list system applied, according to the KPU’s estimate, those who voted for candidates only or both the party and a candidate from the same party totalled 69% of all 104,099,785 votes. The LSI’s exit poll after the 2009 election, in which 3685 respondents were interviewed immediately after they exited the polling stations, showed that 38.1% of respondents reported having voted for a candidate only, while 34.7% voted for both a party and a candidate from the same party, producing a total of around 72.8% who reported using their ballot to mark a preference for individual candidates.¹⁴ In 2014, with a similar electoral system, the KPU estimated that 70% of the 124,972,491 voters marked their ballots for individual candidates and 30% for party only. Clearly, then, over the period of 2004–2014, there has been an increasing trend among voters to vote on the basis of personal candidate preference rather than in response to political party appeals. This increase was particularly evident in the switch from the 2004 semi-open regime list to the 2009 open-list system, but continued in 2014.

The LSI’s post-election survey conducted in June 2014 offered a more detailed picture of voters’ greater inclination to vote for candidates over parties. In the national DPR election, 44.5% of respondents admitted marking the name of ‘candidate only’ and 22.5% voted for ‘the party and a candidate from that party.’ In the provincial and district DPRD elections, the ‘candidate only’ vote was even higher, at 47.5% and 52.1%, respectively (see Table 1.2). It is therefore reasonable to argue that smaller constituencies incentivise candidates to campaign more on the basis of their individual profiles than do larger constituencies. I will elaborate on this point in later chapters.

Table 1.2 Personal votes in national, provincial, and district legislative elections (%): The 2014 legislative election

Obviously, a number of caveats are in order. Although candidates across parties uniformly chased personal votes relying on what Hicken (2007a: 48) termed ‘name and fame,’ this doesn’t necessarily mean that the influence of political ideology and major political figures has disappeared. While trying to expand their electoral bases, most candidates sought to maintain support from their party’s existing religio-political and social constituencies. For example, PKB candidates, especially those running in East Java districts, often produced posters, banners, and stickers that featured a Nahdlatul Ulama—the largest traditionalist Muslim organisation—logo and photos of influential ulama, in an obvious effort to highlight their roots in the traditionalist Islamic community (Fealy, 2014). Similarly, many candidates from PAN deliberately targeted the Muhammadiyah constituency, especially in its strongholds like Aceh and West Sumatra. In contrast, many PDI-P candidates presented themselves as ‘nationalist’ and sent their pluralist messages out to areas inhabited by less religiously observant Muslims and minorities. In short, many candidates presented themselves within the framework of their parties’ popular images (Aspinall and Sukmajati, 2016). Similarly, many candidates still tried to ride the coat-tails of their party’s national leaders. Among presidentialised parties, for example, Gerindra candidates often produced publicity material that promoted their names and photos alongside those of Prabowo Subianto. Even candidates from the more socially rooted PDI-P enthusiastically displayed photographs of Joko Widodo (Jokowi) or Megawati in their advertising materials in 2014.

There is, then, clear evidence that the structural and institutional settings—as well as changes within them—are relevant to the patterns of electoral competition and vote buying in Indonesia. Compared to closed proportional systems, an open-list system provides more opportunities for candidates to determine electoral outcomes—and more incentives to engage in personal campaigning and patronage-based approaches. Declining party attachments and increasing intraparty competition likewise favours clientelist exchanges. The sharp decline of both party identification and party membership over time have further strengthened the move from the party-centric party system of early post-Suharto Indonesia to a more candidate-centred regime, with fundamental consequences for electoral strategizing.

1.6 Design and Methods

In this study, I employ mixed methods by combining quantitative and qualitative approaches. For the study’s main basis, I use survey data to test several hypotheses and conduct statistical analysis. I used qualitative methods to flesh out and explain the survey findings. While the scope of research and analysis focuses primarily on the national level, I also pay considerable attention to the dynamics of vote buying at sub-national levels. This section sets the scene for the interplay between quantitative and qualitative that informs my study and explains how I selected case studies at the sub-national level.

1.6.1 Approaches and Methods

1.6.1.1 Quantitative Approaches

This study draws data from six different surveys as the primary sources of my quantitative research. Most notably, with regard to establishing the level of vote buying in Indonesia’s population, some scholars argue that survey data provide a more sound basis for analysing clientelism than many other approaches because it allows researchers to examine the extent to which vote buying is pervasive and draw inferences about causation (Gonzalez-Ocantos et al., 2012: 203). Survey methods allow us to test competing theories about the determinant factors of clientelist strategies.

But survey methods have their problems. Many scholars argue that quantitative methods are problematic because they miss a lot of the nuance. In addition, using surveys to study vote buying presents structural challenges since many individuals exposed to such practices may not admit to their behaviours (Brusco et al., 2004: 69). In most parts of the world, vote buying isn’t morally legitimate and thus attracts a negative social stigma (Hicken, 2007b; Gonzalez-Ocantos et al., 2012; Corstange, 2012). There is, then, a potential social desirability issue in interviewing and surveying respondents on this topic.

In response to this problem, the first challenge in establishing the extent of vote buying in Indonesia is to minimise potential respondents’ fear of expressing socially undesirable attitudes or admitting to stigmatised behaviours like vote buying. In order to do so, I used a number of survey items of varying degrees of directness (see Chap. 2). My own survey didn’t only rely on a direct question but also used a neighbourhood question by treating respondents as an ‘observer’ to assess the prevalence of vote-buying incidents in their community. More importantly, I employed a list experiment, an increasingly influential mode of quantitative research in studies of clientelism designed to reduce bias in survey questions.

Hence, this study is expected to offer strong empirical evidence of the extent and patterns of vote buying that is generally absent from the more anecdotal accounts that predominate in previous works on Indonesia’s clientelism and electoral politics. This study therefore aims to fill a significant gap in the literature by providing original survey data on vote buying in Indonesia. Using probability-based samples, my surveys include voter-level, candidate-level, and broker-level data.

As an executive director of a Jakarta-based polling organisation, I have extensive experience in organising surveys through face-to-face interviews in Indonesia. Given the sensitive nature of some questions, notably on vote buying, it was essential to secure trust from respondents. At the outset, my interviewers made clear that the survey institute was independent and non-partisan, and that they weren’t affiliated with either any government institution or a particular party. The six surveys that form the quantitative foundation of this study are the following:

• 1. National pre- and post-election surveys of voters: I conducted both pre- and post-election surveys.¹⁵ These surveys asked respondents a variety of questions relating to vote buying, such as how prevalent it was; how effective and costly it was; which kinds of voters were susceptible to such exchanges; and how candidates and brokers monitored compliance. My first pre-election survey was administered during 18–30 January 2014; the second one was conducted from 26 February to 6 March 2014; and the third one was organised from 19 March to 24 March 2014, around two weeks before the election. My post-election survey was run between 22 and 26 April 2014, immediately after the legislative election which was held on 9 April, benefitting from voters’ recent interactions with candidates and brokers. Some measures of vote buying employed in this study were generated from the ‘Money Politics in Southeast Asia’ Project.¹⁶ In addition to these four surveys, I was also able to draw on the results of massive, multi-year national surveys conducted by the Indonesia Survey Institute (LSI), Indikator Politik Indonesia, and Saiful Mujani Research and Consulting’s (SMRC ) in order to incorporate trend data—notably with regard to the aggregate levels of party identification and party membership. I have been deeply involved with LSI and Indikator, and I have been granted permission to use SMRC’s historical data.

• 2. List experiment: Survey findings, however, must be treated with some caution due to the risk of a social desirability bias. To produce more valid estimates of the extent of vote buying and reduce possible errors that may be caused by such bias, this study employed survey experiments, embedded within my two nationally representative surveys of the electorate, by splitting the sample into random halves: a treatment and a control group. The results of the list experiment allow for comparisons with direct and neighbourhood measures to investigate how prevalent vote buying is.

• 3. Pre-election legislative electoral district surveys of voters: I was also able to draw upon massive pre-election electoral district surveys in the lead up to the legislative election (mostly in 2013 and 2014) conducted by Indikator and SMRC in 73 out of 77 electoral districts across Indonesia. The total number of respondents involved in this massive project was 71,940 respondents. In this study, I used these surveys for sub-national case selection (as will be discussed in the next sub-section) and for examining the possible relationship between electoral competitiveness and vote buying in each electoral district (see Chap. 7).

• 4. Pre-election local executive election surveys of voters: My quantitative analysis is also based on extensive empirical data drawn from surveys administered in the lead up to local executive elections by the aforementioned pollsters. These surveys canvassed voter attitudes in the lead up to elections for regional government heads (pilkada) at the provincial and district levels across Indonesia. The total number of surveys used in this study was 1163 with the total number of respondents 725,890. Using several questions relating to vote buying that have been asked since 2006, my study can establish both longitudinal trends and inter-regional variations in levels of vote buying.

• 5. Local post-election surveys of candidates: In addition to drawing on the large national polls and multi-year surveys from local elections, I also designed and organised surveys with elected candidates in four selected provinces (for the selection of these provinces, see the following section). These surveys used face-to-face interviews to ask a set of questions relating to the use of clientelist strategies, recruitment of brokers, targeting strategies, and similar issues. The sample was determined through the multistage random sampling method by grouping the populations in each province based on zones. I divided each province into four different zones, using criteria that varied depending on the geography, history, and conditions of a particular region (see Appendix B). Each zone was a combination of provincial electoral districts. In each zone, a certain regency or city was picked randomly as sample by proportion. In total, the survey consists of 299 randomly selected elected candidates for the DPRD at the provincial and the municipality/district level. As I elaborate in more detail in Appendix B, this rarely used approach enabled me to identify possible patterns in terms of candidate targeting strategies.

• 6. Local post-election survey of brokers: I also conducted surveys through face-to-face interviews with randomly selected brokers in the four targeted provinces. Drawing a probability sample of the candidate survey is unproblematic, given the easy availability of the sampling frame. However, we don’t possess a ready-made sampling frame for low-level operatives from which one could generate a random sample. As explained further in Appendix B, the approach I chose in this context was that a sample was acquired and then estimated based on information received from the randomly selected candidates during the field interview process. Both candidate and broker surveys were supplemented with qualitative observations and interviews—notably on the reasons behind the responses, providing a basis for the valid interpretations of the collected data. The questionnaire I designed aimed to uncover the modes of organisation of brokers, the mechanisms of vote buying (who was targeted, when, how, and why), and the extent to which the recipients of cash handouts voted for the candidate associated with the benefit. Such data allow me to gauge the dynamics of vote buying as they were shaped and driven by grassroots brokers.

1.6.1.2 Sub-national Case Selection

The selection of the four provinces for the sub-national surveys of candidates and brokers was guided primarily by the findings of the wide-ranging electoral district surveys of voters conducted by Indikator and SMRC during the run-up to the 2014 legislative election.¹⁷ I used these surveys to identify variations among provinces in voters’ exposure to vote buying and their levels of party identification (which as already touched on above, I found to be one of the strongest predictors of vote buying). I used these two variables to select cases. My measure of partisanship was based on an additive scale derived from two items: whether respondents felt close to a party and the strength of that feeling. Meanwhile, vote buying also used an additive scale from two items: the extent to which vote buying is reported as acceptable by respondents and the percentage of those who reported to have accepted an offer and voted for the giver.

I used a scatter plot to determine the case selection (Fig. 1.2), selecting one case from each quadrant. I selected Central Java where voters exhibited relatively high levels of both partisanship and vote buying. My second case is North Sulawesi where partisanship is high, but the rate of vote buying is relatively low. The third case is East Java, where voters were more accepting of vote buying, but unlike Central Java, partisan ties were comparatively low. The last case is West Sumatra as a control as the province exhibits relatively low levels of both partisanship and vote buying.

Fig. 1.2

Province selection for MP and broker surveys. Source: Indikator and SMRC’s pre-legislative election surveys in 73 electoral districts during the run-up to the 2014 legislative election

Importantly, these four cases represent the three most populated islands in Indonesia: Sumatra, Java, and Sulawesi. There is also significant variation among the four provinces in critical aspects. As I will discuss later, they vary in the type of political players who are most influential, bases of power, party dominance, socio-economic levels, relative presence of religious leaders, and ethnic composition. These similarities and differences make the four provinces a suitable kaleidoscope through which to view Indonesia’s electoral patterns and politico-cultural composition.

1.6.1.3 Qualitative Approaches

After conducting public opinion and targeted group surveys, I integrated my survey results and my large and complex datasets with the insights gleaned from extensive qualitative fieldwork, which lasted for 13 months. By doing so, I was able to capture—in the context of micro-level analysis—variations, mechanisms, and motivations that characterise the actors and networks through which vote buying was distributed. In order to structure the fieldwork, I chose the case study approach—selecting the four provinces mentioned above, for the reasons already explained. The fieldwork relied on qualitative methods using the same methods across four cases that allowed me to develop in-country comparisons. I spent considerable time in several districts in the four provinces. The criteria for district selection were again primarily guided by statistical findings. For example, in East Java, I travelled to Bangkalan, Sampang, and Pamekasan because, according to my historical data, more people in these areas thought that vote buying was a normal business compared to other regencies.

The qualitative approach relied on four primary tools of investigation: interviews, focus group discussions (FGD), close observation and media analysis, and additional document collection. In conducting the interviews, I used a semi-structured format (Wengraf, 2001; Galletta, 2013). That is, I had certain core topics that I wanted to address in each interview, but I was also flexible in responding to specific cases brought up by each of the informants. Because I had already conducted the main surveys and produced preliminary analysis prior to the bulk of my qualitative research, I was able to zero in on issues of particular interest identified through the quantitative research. For example, given the quantitative result that party loyalists were more likely to be targeted by vote buying, I endeavoured to corroborate and explore this finding in greater detail by asking my interviewees about it. For instance, despite vote buying seeming to target party loyalists, little is known about the definition of loyalty in the Indonesian context, where party-based partisan ties are low by international standards. Thus, I asked my respondents about the meaning of partisanship in the context in which a lot of connections between voters and candidates aren’t mediated by political parties but by informal brokerage networks. And, of course, I had many other questions to ask in terms of the techniques and problems of vote buying, such as: how do politicians engage in vote buying operations? How do brokers identify their voters, and how do they try to persuade them to be responsive if given rewards? And how do they attempt to rely on their ‘loyalists’ if they aren’t given benefits?

In general, the interviews turned out to be highly effective because despite its illegality, many interviewees talked about vote buying openly, with little obvious sense of embarrassment or guilt. I guaranteed that I wouldn’t personally identify them in my research products if they mentioned sensitive matters, and in these circumstances, most were willing to disclose the methods they used to carry out vote buying and explained in detail how they designed their strategies.¹⁸ In the following, I describe my qualitative methods more comprehensively.

• 1. In-depth interviews and informal conversations with politicians: In the four case study areas, I interviewed 24 local candidates. In addition, I also interviewed 42 national politicians from the same four provinces and elsewhere, most of whom ran as candidates in the 2014 elections. Insights from candidates running for the national legislature were important since I had already gained a large amount of data from provincial and district candidates through the local politician surveys. By combining these data with material derived from interviews with national candidates, I could get a sense of how coordination occurs across different levels of competition. Interview questions dealt with the networks and processes through which vote buying was organised, notably questions such as: what is the effect of open-list PR systems on the nature of electoral competition? How do politicians build personal brokerage networks? How do they determine which voters to target? How do they define base voters? How do they align themselves with existing social networks? And how do they monitor their brokers?

• 2. In-depth interviews with brokers: I also chose to interview seasoned brokers. The overall number of such interviews was 28. Given that my broker survey mentioned above wasn’t held in every district of the selected four provinces, I held in-depth interviews with a range of brokers who were representative in terms of district origins. Interview questions probed matters such as what their primary motivations for joining campaign teams were; whether they generally had prior personal contact with candidates; how politicians extract services from brokers; which voters provided the most electoral returns after receiving benefits; how they developed a local following of voters; how payments were presented to voters—using language of gift giving or as a binding transaction; how they defined voter loyalty and base areas; how they diverted resources for their personal benefits; and what measures they used to minimise risk of wastage. In addition, I conducted in-depth interviews with several academics, journalists, non-governmental organisation (NGO) activists, and key bureaucrats about vote buying and its impact on electoral competition in Indonesia.

• 3. Focus group discussions (FGD) with local politicians and political consultants: FGDs are becoming an increasingly popular qualitative research method to collect data from people with similar backgrounds or experiences by using the technique of an organised group interaction focused on a defined topic (Kitzinger, 1994). I organised two focus groups. The first was a group of local politicians from Bukit Tinggi, West Sumatra, held on 23 September 2014. Around 15 successful candidates from various backgrounds attended the event. I chaired the focus group using the Indonesian language and asked participants to reflect on their 2014 campaign, recruitment of brokers, and related matters. The second was a political consultant focus group, conducted on 15 September 2014. Given that the brokers’ survey mainly targeted traditional grassroots brokers, I selected modern-type consultants based in Surabaya to attend the focus group in order to increase the diversity of the broker group I researched. Six participants attended and described their techniques of mobilising voters and the methods they used to monitor campaigns.

• 4. Close observation : This type of collection data is commonly used in qualitative approach aimed to gain a direct, close, and intimate familiarity with a particular social group for period of time, “collect (detailed, comprehensive) field notes, and track systematic patterns to make inferences about social phenomena” (Weiss and Hutchcroft, 2012: 9). During my fieldwork conducted after the 2014 legislative election, around two months before the presidential election, I observed relevant events for the presidential campaign, including political gatherings, campaign rallies, broker meetings, and other important gatherings related to the campaign. Given the observations were conducted after the 2014 legislative election, and vote-buying attempts were more likely to take place in the parliamentary election rather than in the presidential election, as I will show in the next chapter, I didn’t have the opportunity to observe first-hand vote buying efforts. Virtually all local legislative candidates and their brokers I interviewed admitted that such practices were more common during the parliamentary elections. However, as noted above, they didn’t have any objections to discussing the vote-buying strategies they had used during the legislative elections. Additionally, while conducting fieldwork in Central Java and East Java, I also got the strong sense that voters in these two provinces were more tolerant of such election-related bribery than those who lived in West Sumatra.

• 5. Media content analysis and additional document collection: My fieldwork didn’t rely exclusively on relatively common tools such as interviews and direct observation, but I also collected local and national media reports on vote buying and analysed them qualitatively. Among other sources, the media provide a rich collection of information on the occurrence of vote buying, its location, its timing, and its targets. To ensure a broader range of coverage, I also used ‘triangulation of multiple sources’ by taking advantage of the increasing popularity of online news websites in Indonesia as sources of data on clientelist strategies. In addition, I collected a number of official documents related to the subject under investigation. I also collected a large amount of official election statistics by the Indonesian Electoral Commission (KPU ) and demographic data to enrich my quantitative findings.

1.7 The Book Overview

This book presents the dynamics of vote buying in Indonesia’s young democracy, explaining how such exchange is practised and organised, how common and effective is it, and, particularly important, why candidates engage in it while these monetary incentives appear to influence only a limited number of people. My study addresses these central issues in the context of comparative studies of vote buying, arguing that although in relative terms, partisan voters are more likely to be targeted, in absolute terms, vote buying largely happens among undecided voters (given the relatively small number of party loyalists in Indonesia). Regardless of such a substantial amount of leakage, vote buying remains an attractive strategy for many candidates because the 10% range of vote-buying effects on vote choice is high enough to secure victory. Even so, the dynamics of electoral competition that entangle candidates in a prisoner’s dilemma make vote buying as a counter mechanism to neutralise the effects of their rival’s handouts. In addition, vote-buying strategies become relatively more effective in garnering votes than other electoral strategies.

Following this introduction, Chap. 2 discusses a central question of this study. It attempts to answer how extensive vote-buying distributions are in Indonesian elections. Especially with regard to vote buying, I present estimates of these practices using various measures and techniques. This chapter compares the findings from Indonesia with the level of vote buying in other countries. The chapter focuses on vote buying in legislative elections but also pays attention to evidence of extensive vote buying in local executive contests. Utilising a rich vein of data from 2006 to 2015, I identify inter-regional variation in levels of vote buying over time. The chapter concludes that vote buying has become an increasingly prominent feature of Indonesia’s electoral politics at both the national and local levels.

Chapter 3 identifies factors explaining why some individuals are more likely than others to be targeted with vote buying by political operatives. This chapter also provides a complete profile of the typical vote sellers. It rigorously tests the patterns of vote buying based on a survey done after the 2014 legislative election. The findings show that party identification is among the strongest predictors for explaining vote buying. Simply put, the closer the ties of an individual to a political party, the more likely he/she is to be exposed to vote buying.

Chapter 4 then discusses the relationship between party-based partisanship and vote buying in the context of the debate on whether vote buyers are more likely to target swing or party loyalist voters. It discusses the levels of party identification. Given the centrality of the finding that a high degree of party identification in a voter makes him or her more likely to be the target of vote buying, this chapter engages in further tests of this hypothesis with multiple sources of data. First, it reviews evidence from voter-level data which also indicates the greater likelihood of party loyalists being targeted. Moreover, the chapter presents evidence from a unique survey of politicians and brokers which is strongly suggestive of the party loyalist strategy. After interrogating the evidence, the chapter comes to the conclusion that despite the fact that in relative terms party loyalists are more likely to be targeted, in absolute terms vote buying mostly happens among non-partisans—largely because the number of voters with high levels of party identification is small.

Chapter 5 explains the gap between politicians’ intention of capturing party loyalists and the fact that it is undecided voters who most receive benefits. It offers an additional explanation to the conventionally more party-oriented literature by combining an emphasis on the core-voter argument with a stronger focus on candidates’ and brokers’ reliance on personal networks. I call this explanation the ‘personalist loyalist’ argument. I highlight that the concept of loyalty is ambiguous in the Indonesian context, leading many political actors to misidentify (and overestimate the number of) partisan voters. In this regard, institutional and structural contexts matter greatly in shaping the environment in which vote buying can thrive; thus, the chapter begins with a comparison between Indonesia and Latin American countries regarding institutional and contextual factors.

Chapter 6 puts flesh on the bones of my argument about the personal loyalist strategy. It demonstrates how personal networks feed the recruitment and organisation of brokers and help define targeting strategies in settings where party-based partisan ties are weak. The chapter then addresses many of the unanswered questions in the study of electoral clientelism in Indonesia, such as those regarding the demographic profiles of brokers and the logistics of vote buying. Importantly, this chapter argues that the dual-track strategy employed by many candidates, in which they target both party loyalists and persons connected to them through personal networks strategy—not only suffers from misdirected targeting but is also jeopardised by agency loss, that is, leakage of the funds provided by candidates. The discussion shows that, despite reliance on brokerage networks, candidates typically develop weak monitoring and lack enforcement methods, encouraging brokers to engage in rent-seeking behaviour.

Given the widespread leakage and failed targeting, it is crucial to investigate the effectiveness of vote buying. Chapter 7, therefore, begins with discussion of its impacts on both voter turnout and vote choice. It shows that despite targeting strategies being imperfect and despite the unreliability of brokers, vote buying produces greater turnout. This chapter also shows that while receiving money influences the vote choice of ‘only’ approximately 10% of voters, this 10% matters immensely in Indonesia’s highly competitive election settings. It concludes that candidates find vote buying attractive because it serves as an effective mechanism to produce narrow winning margins. The prisoner’s dilemma types of situation also make such practice inevitable, especially in a context where vote buying is relatively more efficient than all the other feasible alternatives.

In the conclusion, I discuss the theoretical and policy implications of the findings and point to an agenda for future research. Overall, the comparative literature on vote buying and turnout buying has emerged in contexts different from that in Indonesia. This makes the dynamics of vote buying, especially its targeting, very distinctive.