Vote Buying in Indonesia The Mechanics of Electoral Bribery

Burhanuddin Muhtadi

Vote Buying in Indonesia

Burhanuddin Muhtadi

# Vote Buying in Indonesia

The Mechanics of Electoral Bribery

Burhanuddin Muhtadi Faculty of Social and Political Sciences Syarif Hidayatullah State Islamic University (UIN) Jakarta, Indonesia

#### ISBN 978-981-13-6778-6 ISBN 978-981-13-6779-3 (eBook) https://doi.org/10.1007/978-981-13-6779-3

© The Editor(s) (if applicable) and The Author(s) 2019 This book is an open access publication

**Open Access** This book is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this book are included in the book's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the book's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Cover Image: © Elena Odareeva / Alamy Stock Photo

This Palgrave Macmillan imprint is published by the registered company Springer Nature Singapore Pte Ltd.

The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

# Acknowledgements

This book is based on a revised version of my dissertation at the Australian National University (ANU). Therefore, I am particularly grateful to my dissertation chair, Prof. Edward Aspinall who played a significant role in the evolution of this project. My special thanks and deepest appreciation also go to Prof. Marcus Mietzner, Prof. Allen Hicken, Prof. William Liddle, and Dr. Ward Berenschot, who each read the entire manuscript and provided insightful comments and useful criticism.

This book could never have been completed without the help and contribution of many people and institutions. First and foremost are my home survey institutes, The Indonesian Survey Institute and Indonesian Political Indicator. My polling organisations not only provided an additional budgetary support to cover the costs of my post-election survey of voters, they also allowed me to insert several questions into a series of nationwide surveys before the 2014 elections.

This project was also enriched by the massive number of electoral district surveys conducted by the Saiful Mujani Research and Consulting (SMRC). Therefore, I would like to express my gratitude to my senior colleague, Prof. Saiful Mujani for introducing me to the tradition of empirical and statistically orientated social science research. I have learned much of what I know about quantitative methods from my bright colleague, Deni Irvani, and I have greatly benefited from discussions with him about experimental design and on how to interpret my results. Of course, I also extend my heartfelt appreciation to all of my interviewees who generously gave of their time in interviews and often lengthy discussions about their political strategies, opinions, and practices.

Finally, I must extend my deepest thanks to my beloved wife, Rahmawati Madanih, and children, Rayhan Adnan Musthafa, Avicenna Ananda Musthafa, and Alexa Shakira Musthafa for their support in completing this project. This book required very costly methods, including the financing of unique, innovative surveys of brokers and politicians. When I was unable to get grants to fully cover all these activities, my wife didn't object to spending our own money. I am also indebted to my parents, H. Muhtadi and the late Hj. Siti Mas'udah. Without their unfailing encouragement and support, I wouldn't have been able to finish this project.

As is the rule in academic writing, I bear sole responsibility for any errors and inconsistencies in this book.

# Contents



# List of Figures






# List of Tables



# Introduction

*"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 decisionmaking 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 multilevel 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 longterm 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

<sup>1</sup>These seven countries included Ghana, Mali, Namibia, South Africa, Benin, Senegal, and Botswana.

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).

2Among 22 countries across the region, according to the LAPOP, the Dominican Republic came out on top, with 22% of respondents claiming that they have been offered material benefits in exchange for their votes. Argentina was second, with 18% of those surveyed reporting having been offered money for their votes either sometimes or often; Panama followed with 17.8%. This big project involved a total of 37,642 individuals selected randomly to represent a population of eligible voters in 22 countries in the Latin American region.

<sup>3</sup> In addition, much has been written about Indonesian voting behaviour from the perspective of sociological context, party identification, and rational choice (e.g. King, 2003; Ananta et al., 2004; Mujani et al., 2012; Liddle and Mujani, 2007).

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 nonparty 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 nonclientelist 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 corevoter 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 preva-

4Among others, James Scott's (1972) "Patron-Client Politics and Political Change in Southeast Asia" is one excellent, classic work on clientelism that emphasised the importance of personal networks. The recent study by Cruz (2014) also finds that those with larger social networks (defined as more friendship and family ties) in the Philippines are more likely to be targeted for vote buying. Similarly, Wantchekon (2003) shows that in Benin, clientelistic goods are often distributed through personal networks.

lent 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

<sup>5</sup>From 2019 onwards, legislative and presidential elections will be held at the same time.

<sup>6</sup>The only exception is Aceh where—as a result of the 2005 Helsinki peace accord—candidates for provincial and district seats are able to compete through local parties.

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

<sup>7</sup> It is important to note, however, that all 'secular' parties reject the term 'secular'—given that the country's state ideology describes Indonesia as multi-religious—and prefer to call themselves 'pluralist' instead.

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.

<sup>8</sup>Pepinsky et al. (2012) argue that Indonesian voters use Islam as a cue when party policy positions aren't available.

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

9The chart is based on the combination of those who reported being active and inactive members of a political party. During legislative elections, candidates typically ask their supporters to register to be party members in order to get benefits they offer such as a free ambulance service and 'grief money' [*uang duka*—i.e. life insurance] for the family should the holder pass away (Ace Hasan Sy, interview, 14 April 2014). Accordingly, party membership often increases during legislative elections.

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


**Table 1.1** Party list position of elected national parliamentarians

Source: Assorted Indonesian Electoral Commission (Komisi Pemilihan Umum, KPU) documents relating to the 2004, 2009, and 2014 legislative elections

10Even some scholars indicated that some parties simply auctioned their winnable positions off to the highest bidder (e.g. Rich, 2013: 75).

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.

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 candidatecentric, 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 copartisans 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

11Another difference is that the electoral law for the 2004 elections stipulated that although voters were allowed to cast a vote for an individual candidate, they also had to vote for the candidate's party. In 2009 and 2014, in contrast, a vote would still be valid even if the voter opted to choose a candidate alone without voting for their party; the personal vote would then be counted for the candidate's party as well.

12Ferry Kurnia, Personal Communication, 3 October 2016.

13For example, Evans (2004: 201) estimated that candidates from PAN and PKS received more individual votes (59% each), compared to PDI-P which 'only' collected 49% of personal votes in total.

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.

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

<sup>14</sup>These data were not initially included in the LSI's exit poll report in 2009 (http://www. lsi.or.id/riset/357/efek-kampanye-terbuka-menjelang-pemilu-legislatif-2009) but were processed for this study.


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

Source: LSI post-election national survey conducted in June 2014

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 subnational 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 voterlevel, 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

15As will be further explained in Appendix A, I organised three pre-legislative election surveys. The first of the pre-election surveys involved 2039 individuals selected randomly to represent the population of eligible voters in Indonesia. The second pre-election survey interviewed 2050 respondents across Indonesia. The last pre-election survey covered 1220 voters across the country. Meanwhile, the number of respondents in the post-election survey was 1210.

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.


16This multi-year project was a four-country study of money politics across Southeast Asia that included Indonesia, Malaysia, Philippines, and Thailand. My thanks to Edward Aspinall and other principal investigators of this project (i.e. Allen Hicken, Meredith Weiss, and Paul Hutchcroft) for allowing me to employ their measures and share the results for the purposes of this study.

between electoral competitiveness and vote buying in each electoral district (see Chap. 7).


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.

17Given many densely populated provinces consist of several electoral districts, I combine all electoral districts in such provinces to get an additive scale of 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.


<sup>18</sup>Aspinall and Sukmajati (2016) along with approximately 50 researchers involved in 'Money Politics and Clientelism' project also had similar experiences. Many candidates they interviewed were relatively open when discussing how the payments were delivered to the recipients.

activists, and key bureaucrats about vote buying and its impact on electoral competition in Indonesia.


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.

#### Bibliography

Books, Journal Articles, and Unpublished Papers


Ananta, A., Arifin, E.N., and Suryadinata, L. (2004) Indonesian Electoral Behavior: A Statistical Perspective. Singapore: ISEAS.


———. (2007b) How Effective are Institutional Reforms? In: F.C. Schaffer (ed) Elections for Sale: The Causes and Consequences of Vote Buying. Colorado: Lynne Rienner Publisher, Inc.


———. (2007) Clientelism, Credibility, and the Policy Choices of Young Democracies. American Journal of Political Science 51(4): 804–821.


———. (1969) Corruption, Machine Politics, and Political Change. American Political Science Review 43:1142–1158.


———. (2004) The 2004 Indonesian Elections: How the System Works and What the Parties Stand for. Canberra: Centre for Democratic Institutions.


Tomsa, D. (2008) Party Politics and Democratization in Indonesia: Golkar in the Post-Suharto Era. London and New York: Routledge.

———. (2014) Party System Fragmentation in Indonesia: The Sub-national Dimension. Journal of East Asian Studies 14: 249–278.


**Open Access** This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

# The Prevalence of Vote Buying in Indonesia: Building an Index

How prevalent is patronage distribution in Indonesia? Although there has been a burst of scholarly and non-scholarly writings on the topic in the last few years,1 little is known about how many voters actually receive material incentives from politicians. This chapter offers a systematic answer to fundamental questions about the intensity of money politics in Indonesia that have vexed scholars for many years. It involves a complex study of patronage politics in Indonesian elections, which draws predominantly on surveys.2

1Among others, Choi (2007), Aspinall (2014), Allen (2015), Aspinall and As'ad (2015), Aspinall and Sukmajati (2016), Aspinall et al. (2017), Amick (2016), and Tawakkal et al. (2017).

<sup>2</sup> I used a number of nationwide surveys in this chapter, including an April 2009 survey by The Indonesian Survey Institute (LSI) and a December 2013 survey by Saiful Mujani Research and Consulting (SMRC). The LSI survey took place from 20 to 27 April 2009, approximately one week after the legislative election. The sample was 2000 voting-age adults with a margin of error of ±2.4% at 95% confidence level. Meanwhile, the SMRC survey interviewed 1210 respondents, who, like the respondents in the LSI poll, were selected with multistage random sampling, with an estimated margin of error of ±2.9% at 95% confidence level. This chapter also draws from a pre-election survey conducted during 19–24 March 2014. The number in the sample was 1220 voting-age adults who were selected with multistage random sampling proportionally distributed over the 34 provinces. The margin of error was around ±2.9% at 95% significance level. However, this chapter draws primarily from my post-legislative election survey conducted during 22–26 April 2014, around two weeks after the legislative election. I timed the surveys to take advantage of a simultaneous national and sub-national parliamentary election held on 9 April 2014, which would mean that citizens' recent interactions with parties, candidates, and brokers were fresh in their minds. In addition to these national surveys, I also utilised a large amount of data drawn from 963 local surveys conducted by LSI and Indikator Politik along with SMRC from 2006 to 2015.

© The Author(s) 2019 45 B. Muhtadi, *Vote Buying in Indonesia*, https://doi.org/10.1007/978-981-13-6779-3\_2

In this chapter, to measure the intensity of vote buying, I use a votebuying index as the main dependent variable of this study. I argue that vote buying is central to election campaigns in Indonesia. Drawing from a wide range of survey methods, I estimate that the proportion of voters participating in vote buying during the 2014 legislative election was between 25% and 33% (depending on the method and specific question used in the survey). For comparative purposes, I relate this figure of selfreported vote buying with rates from other countries in the world. It turns out that the range between 25% and 33% is comparatively high by international standards, with Indonesia's level of vote buying being the third largest in the world (at least as measured in surveys within the last decade). As vote buying is illegal in most countries, including Indonesia, it is plausible to suspect that respondents directly asked about such exchanges are reluctant to provide truthful answers. However, despite this presumed desirability bias, my findings show that the list-experiment and the straightforward survey questions result in consistent estimates of the aggregate levels of vote buying. The chapter discusses this question extensively.

Finally, although this study puts greater weight on vote buying in national parliamentary elections, it also offers new insights into the underexplored nature of such practices in different election settings. In the last part of this chapter, I present historical survey data, owned by my home institutions (LSI and Indikator Politik Indonesia), to identify interregional variation in levels of vote buying in executive elections at the local level. This chapter concludes that vote buying isn't only widespread during national legislative elections, but that it has also become a central feature of local executive elections, where the prize is the position of district head, mayor, or governor.

### 2.1 Dimensions and Measures

To operationalise vote buying as the dependent variable of this study, I created a series of questions about this individualised form of patronage distribution and asked subjects to respond on a various point Likert-type scale to capture attitudes and opinions with a range of answer choices. Given that vote buying is typically associated with negative stigma and thus is associated with social desirability bias, my research follows the methods developed by Brusco and her collaborators (2004: 69) to approach the targeted issue from several different directions. In the postelection survey (the one conducted from 22 to 26 April 2014), the questions used seven measures of vote buying with different degrees of directness. This was done in the expectation that the response to any single question wouldn't provide a completely accurate picture of such exchanges. Further, I categorised these seven measures into two dimensions: vote-buying operations in the nationwide legislative elections and vote-buying attempts in the national and local executive elections.

In the survey, the dimension of vote buying in the parliamentary elections consists of four measures. The first is a four-point scale of all experiences with vote buying in legislative elections, without concretely mentioning the 2014 elections in the question (hereafter I referred to this a four-point scale of all vote-buying experience in legislative elections). The question asked: "During the last couple of years, related to the legislative election campaign of national parliamentary (DPR) candidates, how often have candidates or success team members offered you food, household items, and/or other goods in order to influence your vote in the election?" I asked respondents to respond on a four-point Likert-type scale (very often, quite often, rarely, and never), which I then recoded into two categories (very often, quite often, and rarely = 1, rare and never = 0). I worded the question in this way to gather all experiences of vote buying, including in the currently ongoing campaign. The strategy of not explicitly mentioning the date was intended to capture greater vote-buying reports. Given the survey was conducted when the election had just been held, I wanted to avoid making respondents feel they were being 'interrogated' on what they had just done in the 2014 election, remembering the vote buying is actually illegal in Indonesia. In addition, as noted above, this measure used a four-point scale to increase variation in this variable of interest and was expected to collect greater vote-buying reports than a simple 'yes-no' question.

Similarly, my second measure of vote buying in legislative elections didn't concretely mention the 2014 election. The prompt was as follows: "These situations sometimes occur during every election in Indonesia. Have you ever experienced these situations below?" The relevant prompt was "Being offered money or goods in order to vote for a certain political party/DPR candidate." The possible responses were 'no,' 'yes, only once or twice,' to 'yes, several times,' and they were recoded to be a two-point scale (yes, only once/twice and yes, several times = 1, no = 0). The main difference to the question above is that while the first measure used a fourpoint scale, the second used a three-point scale of vote-buying experience (hereafter referred to as a three-point scale of all vote-buying experience in legislative elections). We should expect a different outcome since different response options may generate different results. Additionally, the second measure was mixed with a number of other exposures to vote-buying offers in the different election settings that voters could point to, as will be discussed later.

The last two measures of vote buying in legislative elections are on a two-point scale, comprising a pair of questions that explicitly asked about these practices in the 2014 legislative elections. A key difference to the questions above is that respondents might have felt compelled to provide misleading answers in response to questions that directly asked about vote buying in 2014, because this event was still fresh in voters' minds when the post-election survey was conducted and the questions were more specific. Though the question is appealing because it asks directly about the level of vote buying in 2014, it also might drive some respondents to respond in ways that don't reflect their actual behaviour. The question doesn't allow respondents to disguise their answers in terms of past election events. Further, these last two measures are dichotomous scales that are different from the questions above, being a 'yes-no' question which provides a clearer, binary response, but in which bias is more likely.

In this study, the yes-no questions that explicitly mentioned the 2014 elections asked about individual and neighbourhood vote-buying rates. Hence, the third measure inquired about vote buying directed at respondents in 2014 (hereafter referred to as a two-point scale of individual vote buying in the 2014 legislative election). The question reads: "During the run-up to the April 9th 2014 legislative election, did candidates or success team members offer you money, food, household items, and/or other goods (excluding propaganda hats, shirts, and posters)?" The fourth focuses on vote-buying offers witnessed by respondents in their neighbourhood (hereafter referred to as a two-point scale of neighbourhood vote buying in the 2014 elections). The question asked: "During the runup to the April 9th 2014 election, did you observe candidates or success team members offering people in your neighbourhood money, food, household items, and/or other goods (excluding propaganda hats, shirts, and posters)?" These yes-no questions were recoded to scale from 0 (indicating those who responded negatively to both such offers in 2014) to 1 (indicating those who responded positively to both questions).3

<sup>3</sup>To measure vote buying, I prefer to use 'being offered' rather than 'received' material benefits. Of course, we cannot determine whether voters actually accepted electoral bribes.

Meanwhile, three questions focused on vote-buying attempts in national and sub-national executive elections, as a comparison with the above second measure of general vote-buying experiences in legislative elections. These measures of vote buying in the executive elections are on a three-point scale. The first question reads: "Have you ever experienced being offered money or goods in order to vote for certain presidential/ vice presidential candidates?" (hereafter referred to as a three-point scale of vote buying in presidential elections). The second question asks: "Have you ever experienced being offered money or goods in order to vote for a certain gubernatorial candidate?" (hereafter referred to as a three-point scale of vote buying in governor elections). The third question was: "Have you ever experienced being offered money or goods in order to vote for a certain regent/mayoral candidate?" (hereafter referred to as a three-point scale of vote buying in regency elections). In each case, possible responses were 'no,' 'yes, only once/twice,' and 'yes, several times.' These answers were subsequently recoded to be positioned on a two-point scale ('yes, only once/twice' and 'yes, several times' = 1, 'no' = 0).

In my analysis of the data, I combined those seven measures described above into an aggregate index of vote buying. As indicated above, to facilitate substantive interpretation of these findings, I recoded each of the vote-buying items from their original scale into dichotomous scales that took a value of 1 if the respondent was exposed to vote-buying attempts or experienced being offered such electoral bribes.

#### 2.2 How Prevalent Is Vote Buying?

How widespread is vote buying in Indonesia judged from the measures I have chosen? In this section, I present the main descriptive results of my findings on vote buying, divided into two dimensions: legislative elections and executive elections.

To begin, I demonstrate the findings based on the first dimension of vote buying generated from responses to a battery of four questions that asked about respondents' experiences of vote-buying attempts in legislative elections. Using a four-point scale, as explained above, responses to

But such a question, as Jensen and Justesen (2014: 224) argued, allows us to measure which voters political machines tend to target. By using the word 'being offered,' I intended to avoid social desirability biases that may occur when asking questions about respondents' own behaviour.

the first question suggest that 33.1% of respondents reported that they had been offered electoral incentives 'very often,' 'quite often,' or 'rarely.' Although a substantial proportion of the sample (58%) reported having not been targeted at all, around one-third of the respondents admitting such an experience is a relatively high result (see international comparisons in Sect. 2.4 below), suggesting that such practices have become a conventional strategy in electoral campaigning in Indonesia.

The second measure of vote buying in parliamentary elections produced similar results. Using a three-point continuous scale, this indicated that 29% of respondents reported being offered cash or goods 'once or twice' or 'several times' during legislative elections. Using similar wording, surveys taken between 2009 and 2014 showed an increase in the rate of vote-buying offers during legislative elections.4 In the 2009 legislative elections, only 11.2% of respondents admitted being targeted for vote buying. The incidence increased in December 2013 (several months before the 2014 legislative elections) to 20.6% before rising again to 29% in April 2014 (Fig. 2.1).

Likewise, a descriptive overview of the third and the fourth measures of vote buying in legislative elections, which point to direct individual and neighbourhood vote buying in the 2014 legislative elections, result in consistent findings. Using a dichotomous scale, overall, 25% admitted having been personally targeted by such exchanges during the 2014 campaigns (Fig. 2.2). The fourth measure, the neighbourhood measure, showed a higher result of 28.9% (Fig. 2.2).

Although the difference between the various measures is relatively small, ranging from 25% to 33% (see the first four rows of Table 2.1), one might question about the reasons for these differing results. One probable explanation is that different survey response scales may result in different responses. As noted earlier, individual vote buying in the 2014 legislative election was framed in a straightforward question, asking for a yes-no response, while the first question on overall vote buying in legislative elections was on a four-point Likert-type scale in which those who responded affirmatively ('very often,' 'quite often,' and 'rarely') were categorised into

<sup>4</sup> Some of the April 2009 survey findings were reported at the LSI official website (http:// www.lsi.or.id/riset/370/Diskusi%20LSI%20Kualitas%20Pemilu%202009). But the data cited in this study weren't included in the report and were processed for purposes of this analysis. Similarly, the data quoted from the SMRC 2013 survey were initially not included in that organisation's official report, but I was able to analyse them for this study.

**Fig. 2.1** Increasing rates of being targeted for vote buying, 2009–2014 (%). Source: The April 2009 data was drawn from LSI's survey, the December 2013 data was taken from SMRC's survey, and the April 2014 numbers were drawn from my post-election survey

**Fig. 2.2** Individual and neighbourhood vote buying in 2014 (%). Source: My post-election survey, 22–26 April 2014


**Table 2.1** Relative frequency of responses to various measures of vote buying

one category as recipients of vote-buying offers. Similarly, the second question which asked all vote-buying experiences was on a three-point scale, providing an option for respondents to report rarity (i.e. 'once or twice') regarding the frequency of being offered vote buying. Perhaps, these greater ranges of responses captured more experiences than the sharply dichotomous scale. But why did the two-point scale used for neighbourhood estimates also result in higher numbers? As Kramon (2013: 70) points out, the neighbourhood measure is less obtrusive than direct, individual measures, because it asks respondents about others' actions. Another reason that neighbourhood vote-buying totals might be larger than selfreported vote-buying totals is that if such practice is open and common, then a single act of vote buying may be observed by several individuals.

Another possible interpretation is that asking about general votebuying experiences without specifically mentioning the 2014 elections would likely generate higher positive answers than questions that concretely point to 2014. As noted above, mentioning the concrete event can make a psychological difference in that voters may feel pressured to provide truthful answers in response to sensitive questions about a recent act which, in a context like Indonesia, is unlawful. Again, given the timing of the survey, even when posed a question on vote buying that didn't concretely mention 2014, respondents would have included their 2014 experience in their answer. In fact, it was probably their main reference point. In short, mention of the 2014 elections apparently biased the responses of some respondents who were reluctant to report their recent illicit behaviours.

However, as will be explained in the following section, the results of the list-experiment, which is an increasingly influential technique aimed to minimise response bias in survey questions, show that the estimated percentage of people receiving money or gifts was around 27.4%. Thus, the 25% of those who experienced vote-buying offers based on the estimate from a dichotomous scale of the vote-buying experience in the 2014 election sounds about right. However, in order to give a sound estimate that reflects a number of different measures, this study suggests that between a quarter and a third of respondents reported having been targeted for vote buying in nationwide legislative elections.

Considering that the percentage difference in the final results between the first-placed and second-placed parties, PDI-P and Golkar, was only 4.2%, this incidence level of vote buying is significant. Moreover, the percentage difference between PDI-P in the first place and Hanura in the tenth place was just 13.69%—well below the vote-buying levels (The Election Commission [KPU], 2014). And as I will show later, the margin of victory for individual candidates was considerably smaller than the margin between parties, making the prevalence of vote buying even more consequential.

Compared to the pre-election survey carried out in March 2014, which captured only half the campaign period, vote-buying incidents increased significantly from 10.7%. This increase suggests that vote buying was more concentrated in the days or hours before election day, or on election day itself. In the Indonesian context, this phenomenon is popularly known as *serangan fajar* (dawn attack), a term that reflects the fact that material benefits (especially cash) to compensate for votes are sometimes distributed just after the dawn prayer on voting day. Candidates and brokers assume that the closer to voting day they hand things out, the more effective vote buying is in shaping electoral behaviour. These issues, however, aren't unique to Indonesia. The pace of vote buying accelerates as the election date approaches in other countries too, such as in Taiwan (Wang and Kurzman, 2007) or the Philippines (Pulse Asia, 2013).

In order to provide a complete picture of vote-buying practices, those who reported having been offered benefits in the 2014 legislative elections were also asked to name the party or the party of the candidate involved in the transaction. Respondents were allowed to mention more than one party, because they might have been approached by candidates from different political parties (see Chap. 7). In their answers, respondents identified all political parties which ran in the 2014 elections, with varying degrees of involvement. This survey finding corresponds with my qualitative observations. "No party was innocent of vote trading. Even candidates from Islamic parties engaged in such practices," said a one-time national parliament member. "I dare you," he continued, "cut off my finger if there is a single MP in Indonesia today who gets elected without buying votes!" (Interview, 20 April 2014). Although vote buying was also common among Islamic parties, the big four vote buyers were non-Islamic in character: of voters who admitted being offered money or a gift, 32.2% mentioned having received such offers from candidates from Golkar, followed by PDI-P (26.5%), Gerindra (25%), and the Democratic Party (18.4%). Meanwhile, the most common item offered to voters was money (75.5% of all reported attempts), food products (12.8%) such as rice, sugar, and noodles, and household items (11.4%).

As noted earlier, the first three measures consisted of straightforward survey questions about whether individuals encountered offers of material rewards in return for their votes. The fourth measure, however, asked respondents to report whether they observed other people in their neighbourhood or village experiencing such encounters in 2014. Arguably, this question may be less accurate since it is based on perceptions, which can be accurate or not. Nevertheless, asking questions about illegal practices by treating respondents as an 'observer' is less vulnerable to response bias because the question doesn't require respondents to report on their own potentially illicit behaviour (Kramon, 2013: 70). The results of this fourth measure showed that 28.9% of respondents had witnessed or knew that such practices occurred in areas where they lived (see Fig. 2.2). These figures are only slightly larger than responses to the direct question (the third measure), in which 25% of respondents admitted being offered a vote-buying exchange in the 2014 election. The gap between the two is within the margin of error used in this survey and is therefore statistically insignificant.

The results from both the neighbourhood and individual measures are also consistent in terms of the perpetrators and items of vote buying. Among those who reported that their neighbourhood was being targeted, Golkar again topped the list of most frequently mentioned vote buyers (35.5%), followed by PDI-P (30.6%), Gerindra (26.8%), and the Democratic Party (20.6%), repeating the order in the individual vote-buying measure. Similarly, as among respondents who individually observed vote buying, all political parties were mentioned by at least some respondents as engaging in vote-buying efforts in the neighbourhood measure. Money, food, and household items were also again the most frequent goods reported as being distributed in exchange for votes. Of those who witnessed their neighbourhood as having been targeted, 78.9% reported that people in their locations were given cash handouts, followed by food (12.8%), household goods (11.4%), and other items (14.8%). Note that the respondents were allowed to provide multiple answers.

In order to make a direct comparison of different election seasons, my 2014 voter survey also asked a number of questions about the 2009 legislative election, using wording nearly identical to the individual and neighbourhood items. The results confirm my earlier finding (based on 2009 and 2013 surveys) that vote buying was more pervasive in 2014 than in 2009. Only 10.1% of respondents reported having individually been the target of such handouts in 2009. In terms of the background of vote buyers, the vote market was again dominated by non-Islamic parties such as Golkar, PDI-P, and Democratic Party. The most common items distributed to voters in 2009 were also cash payments, foodstuffs, and household goods. When asked about the extent to which vote buying in 2009 happened in their neighbourhood, though statistically insignificant, affirmative responses to this inquiry were a little bit higher (12.9%) than responses to the questions asked about the individual experiences of survey respondents (10.1%).

Having discussed vote buying in the legislative elections, we are now in a position to explore the second dimension of this form of patronage exchange: vote buying in executive elections at the national and local level. Figure 2.3 demonstrates survey estimates of the proportion of voters who experienced being offered electoral incentives in such elections. In terms of presidential elections, my survey in April 2014 (around three months before the presidential elections)5 showed that only 8.1% of those surveyed reported having received vote-buying offers either 'once or twice' or 'several times.'6 In addition, according to the same April 2014 survey, vote-buying reports in the regency and gubernatorial elections were relatively similar, standing at 13.6% and 11.3%, respectively. It is plausible to

<sup>5</sup>Unfortunately, this question wasn't asked in our surveys that were closer to the 2014 presidential elections, or after it.

<sup>6</sup>Although this figure was much lower than in legislative elections, based on the LSI postelection survey in April 2009, there has been a significant increase compared with 2009 when only 3.2% reported an acquaintance being offered such incentives.

**Fig. 2.3** Vote-buying offers at the national and local executive elections (%). Source: My post-election survey, 22–26 April 2014

conclude, therefore, that vote-buying attempts were more prevalent during parliamentary elections relative to presidential and sub-national executive elections (I will present additional data from an historical dataset on sub-national elections that confirms this finding at the end of this chapter).

These results are in line with the research of other scholars who argue that constituency size matters in explaining the different levels of vote buying in multiple election settings. Stokes (2007: 86), for example, argues that vote buying decreases as constituency size grows, and vice versa. When the constituency is very large, as in presidential elections, it is difficult for politicians to buy the amount of votes necessary to make a nationwide difference. Presidential candidates instead rely more on the national media in attempting to reach out to a broad audience. By contrast, in direct local elections in smaller rural districts (*kabupaten*) and urban municipalities (*kota*), the cost of vote buying is less expensive and the practical challenges less daunting, thereby creating an incentive for candidates to engage in such practices. In national parliamentary elections—which comprise the national-level DPR, or People's Legislative Council; 34 province-level DPRDs (Regional Legislative Councils); and DPRDs in 508 *kabupaten*/*kota*—vote buying is also widespread because the various electoral constituencies are manageable in size (being concomitantly smaller than the region-wide electorates in the relevant executive government head elections), and because candidates at different levels can coordinate their vote-buying efforts. Aspinall and Sukmadjati highlight the point that since "*kabupaten* and *kota* electoral districts are nested inside provincial electoral districts, which are in turn nested inside national electoral districts…. it is relatively easy for candidates running at different levels to coordinate their efforts in the hope of maximising their individual chances of success" (2016: 14–15).7 This, in turn, increases the likelihood of vote buying in the legislative elections (see Chap. 6).

Another plausible explanation for why vote-buying efforts are more common in legislative elections is that the number of candidates running for national, provincial, and district parliaments is much larger than that of competitors in presidential and regional head elections (see Mujani et al., 2012: 98). The number of seats contested in the 2014 legislative elections was 19,699, which were situated in the national, provincial, and district legislative assemblies. A total of 6608 candidates ran for the 560 seats in the DPR, around 10,000 candidates competed at the provincial level, and hundreds of thousands of candidates stood at *kabupaten*/*kota* level. In a single electoral district, there could be as many as 144 candidates competing at each level of legislative elections. In addition to the DPR and DPRDs, candidates also competed for the 136-seat Regional Representative Council (DPD).8 In the 2014 presidential elections, by contrast, there were only two presidential and vice presidential candidates in the race. Similarly, since 2005 (when direct local elections were introduced), there have never been more than 11 pairs of candidates in such a contest. Thus, the relatively limited number of candidates running for either presidential or local elections reduces the incentive for them to engage in vote buying if compared to their legislative counterparts running in massive, more competitive elections.

<sup>7</sup>The exception to this general rule is Java where national and provincial electoral districts share the same boundaries. For further discussion, see Aspinall and Sukmajati (2016).

<sup>8</sup>Detik.com, "200 Ribu Caleg yang Berebut 19 Ribu Kursi di 2014," 9 January 2014, available at https://news.detik.com/berita/2462640/200-ribu-caleg-yang-berebut-19 ribu-kursi-di-2014, accessed 3 March 2016.

# 2.3 The Vote-Buying Index

Having discussed the descriptive results, I can now develop an additive index of 'vote buying,' based on responses to the battery of seven questions summarised in the preceding section. The Cronbach's alpha coefficient for the variables that make up the vote-buying index is 0.841, meaning that the variables have a high degree of internal consistency, and this suggests that the items in the test are highly correlated. As noted above, I differentiate vote buying into two dimensions: vote buying in parliamentary elections and vote buying in the presidential and local executive elections.

In order to establish that vote buying comprises these two dimensions, I use a confirmatory factor analysis. As shown in Table 2.2, the variables that make up the vote-buying index yield two components: (1) the four items related to vote buying in parliamentary elections constitute a single factor, and (2) the three items related to vote buying in the national and local executive elections produce another dimension. In other words, the variables that relate to vote-buying experiences in executive elections are


**Table 2.2** Factor analysis of vote buying (Rotation method: Varimax with Kaiser normalisation)

Extraction method: Maximum likelihood

empirically distinct from those that for legislative elections. This factor analysis reinforces the notion that the categorisation of vote buying based on the two distinct dimensions above is valid.

Hence, throughout this study, I apply the index of vote buying in statistical equations, including cross-tabulation, bivariate and multivariate analyses. I start this exercise by index-scoring the two dimensions of vote buying. First, the mean score of vote buying in parliamentary elections derived from the four measures mentioned above—which vary from 0 (didn't engage) to 1 (engaged) was 0.3145. Importantly, the bivariate statistical analysis reveals the correlation pattern between all vote-buying variables in the legislative elections (Table 2.3).

Second, the overall score of vote-buying offers in national and local executive elections—normalised to vary between 0 (didn't engage) and 1 (engaged)—was 'only' 0.1121. I also conducted a bivariate test with a Pearson correlation to look at the relationship between all three primary variables that constitute the dimension of vote buying in elections for presidential and regional heads (Table 2.4). Based on this bivariate test, the relationship between vote buying in presidential elections, vote-buying offers in the gubernatorial elections, and vote buying in the regency/mayoral elections is substantial and statistically significant.


**Table 2.3** Correlation between vote-buying variables in parliamentary elections

\*\*\*Correlation is significant at the 0.001 level (two-tailed)


**Table 2.4** Correlation between vote-buying variables in the national and local executive elections

\*\*\*Correlation is significant at the 0.001 level (two-tailed)

Recall again that the overall score of the vote-buying index is an additive scale based on responses to a battery of seven measures that make up the two dimensions of vote buying described above. The scale runs from 0 (not exposed to vote buying) to 1 (exposed). The average of the votebuying index is relatively high, with a score of 0.2103. There is, however, substantial component variation. At the low end, people who were exposed to vote-buying attempts during an election for national and regional government heads only make up a modest proportion of the total electorate, with a score of 0.1121. In particular, people were less likely to have experience of vote buying in presidential elections, with a score of 0.0841. At the top end, people were more likely to be targeted for vote buying in parliamentary elections, with a score of 0.3145. Therefore, as with the above descriptive analysis of the measures, it is safe to conclude from the index that among the two different election settings, vote buying is more common during parliamentary elections than during national and local executive elections. The reasons behind these different rates of vote buying have been already explained above: the effect of constituency size and the differences in competitiveness between these two types of elections.

Although the average of vote-buying occurrence in parliamentary elections is higher than in executive elections, the bivariate model suggests that these two dimensions of vote buying have a positive correlation and are statistically significant in the expected direction (Table 2.5). It can be inferred, therefore, that those who have traditionally benefited from clientelist exchanges in parliamentary elections are more likely to be targeted for vote buying in the national and sub-national executive elections as well. The correlation also exists the other way around. Those


**Table 2.5** Correlation (Pearson's *r*) between vote-buying variables in the parliamentary elections and executive elections

\*\*\*Correlation is significant at the 0.001 level (two-tailed)

being exposed to vote-buying attempts in presidential and regional head elections are more likely to be the target of such exchanges in legislative elections.

What does this correlation suggest? Concretely, it implies that political machines are more likely over time to direct their vote-buying efforts more heavily towards a group of voters for whom vote buying is a more common, repeat experience. This finding is in line with the concept of base voters, according to which loyalist voters are locked into long-term political relationships with parties or politicians based on patronage ties (see Chap. 5). In many cases, political agents (or brokers) assume that such voters are more responsive to clientelist exchanges, which in turn triggers an increase in vote buying targeting this group. Furthermore, as I elaborate in more detail in the following chapters, the logic of various elements of Indonesia's institutional framework, such as optional voting, open-list proportional system, and ballot secrecy, create strong incentives for candidates to target such voters because they are thought of as being more reciprocal, and thus a more predictable source of votes.

#### 2.4 International Comparability

I have explained at length the ubiquity of vote buying in Indonesian elections. To put these numbers into perspective, it is important to compare the rate of self-reported vote buying, especially in parliamentary elections, with rates from other countries in the world. As indicated earlier, I used a number of questions in my surveys that are partially derived from and are comparable to similar questions that have been asked elsewhere. This allows me to make direct comparisons with vote-buying levels in other countries, as presented in Table 2.6.


**Table 2.6** Estimated proportion of direct vote buying by country

Sources: The Latin American Public Opinion Project (LAPOP) Americas Barometer 2010 and the Afrobarometer Round 5, 2011–2012. The data from Malaysia was taken from Meredith Weiss, "General Election 2013 Survey Results" (2013), while the rate of vote buying from the Philippines was taken from Pulse Asia, "On the 2013 Elections: Observations and Select Survey Results" (2013) (The data from Malaysia and Philippines were presented at "Workshop of Money Politics, Patronage and Electoral Dynamics," Yogyakarta, 13–15 December 2013)

a Derived from the first measure of all vote-buying experience in legislative elections

To that end, I use an estimate based on my first measure of all votebuying experience (it will be recalled that this was responses to the question: "During the last couple of years, related to the legislative election campaign of national parliamentary (DPR) candidates, how often have candidates or success team members offered you food, household items, and/or other goods in order to influence your vote in the election?"). This question was similar to questions asked by both LAPOP Americas Barometer and Afrobarometer. The data on vote-buying levels in Latin American and African countries presented in Table 2.6 are all taken from the two polling associations. While the wording of questions varied slightly in some of the studies, the questions were overall congruent. The LAPOP worded the question as follows: "In recent years and thinking about election campaigns, has a candidate or someone from a political party offered you something like a favour, food, or any other benefits or thing in return for your vote?" Respondents could indicate that they 'never,' 'sometimes,' or 'often' have been offered electoral incentives in return for their vote. Responses were coded 'yes,' if they reported 'often' or 'sometimes' having been the target of such practice.

The Afrobarometer asked the following question: "And during the last national election in [20xx], how often, if ever, did a candidate or someone from a political party offer you something, like food or a gift or money, in return for your vote?" Respondents responded on a five-point Likert-type scale, with responses ranging from 'never,' 'once or twice,' 'a few times,' 'often,' and 'no experience in the past year.' The main difference is that I used a four-point scale in the first measure, while the LAPOP Americas Barometer and Afrobarometer employed a three-point scale and five-point scale, respectively. In the case of Malaysia and the Philippines, the frequency of vote buying in both countries was based on a dichotomous scale: whether or not respondents had been offered material benefits during the current/most recent elections.9

Turning to the comparison of vote-buying levels, Table 2.6 shows that there is a wide variation in the level of self-reported vote-buying transactions across in the continents of Asia, Latin America, and Africa, in surveys taken over the last decade. The average level of vote buying around the world is 14.22%, with some countries scoring well below the global aver-

<sup>9</sup>The question reads: "During the run-up to the 2013 elections, did any political party offer you money, food, household items, and/or other goods (excluding propaganda hats, shirts, and posters)?"

age and some significantly above it. Vote-buying intensity is greatest in the cases of Uganda (41%), Benin (37%), Indonesia (33%), Kenya (32%), Liberia (28%), Swaziland (27%), Mali (26%) and Niger (24%). Remarkably, the level of vote buying in Indonesia, when we use the higher estimate (33%), is more than double the global average.

In contrast, vote-buying incidents were virtually non-existent in Lesotho (2%), Mauritius (1%), and Tunisia (1%). In general, consistent with the expectations from prior findings, offers of vote buying in many African and Asian countries are reported in higher numbers than in Latin America (Schaffer, 2007). Confirming this pattern, the study of De Jonge (2015)—which analysed survey data from ten elections in eight Latin American countries—also found that vote-buying incidence wasn't as high as in other parts of the world.10 In contrast, it is clear that vote buying in Indonesia is comparatively prevalent, and that such practices are central in the country's electoral politics.

### 2.5 Little Social Desirability Bias

Given that these data are entirely based on surveys, we must ask: how valid are estimates of vote buying based on direct individual measures established through polls? As noted above, vote buying is illegal in most countries, including Indonesia, and usually linked to a negative social stigma (Gonzalez-Ocantos et al., 2012; Hicken, 2007). Therefore, some scholars have expressed suspicions that respondents directly asked about such practices are unlikely to report accurate information due to the problem of social desirability (Gallego and Wantchekon, 2012; Corstange, 2012). In order to address this concern, survey-based experiments are suitable instruments to evaluate the credibility of individual survey items and consolidate their overall validity. This is particularly the case for clientelism research, which has attracted renewed scholarly interest in recent years.

In the following, therefore, I present my findings on the estimated proportion of people receiving material benefits for their vote based on a listexperiment, embedded within two nationally representative surveys, conducted before and after the 2014 legislative elections. There were two underlying assumptions for conducting the list-experiment. First, we must assume that respondents give truthful answers to the sensitive item (Blair

<sup>10</sup>For my comparative overview, I chose the LAPOP Americas Barometer (2010) data which provides a higher aggregate estimate of such incentives in the case of South American countries—because it used a more comparable direct survey item.

and Imai, 2012: 56). Second, we must assume that "the inclusion of a sensitive item has no effect on respondents' answers to control items" (Blair and Imai, 2012: 51).11

First of all, the list-experiment participants were divided into four random halves: one control group and three treatment groups.12 Interviewers read the same question for each group. Respondents in each group were also provided with a show card for their response choices, which differed by group only in the number of response categories. These items were designed together with the project leaders of 'Money Politics in Southeast Asia,' while considering floor and ceiling effects that must be avoided.13 The opening statement in the first list-experiment was the following question:

I am going to read you various activities, and I would like for you to tell me if they have been carried out by candidates or team success members during the run-up to the April 9th 2014 election? Please don't tell me which ones, only how many? (one, two, or three campaign activities)

For the control group, I listed the following campaign activities:


<sup>11</sup> See also, Eric Kramon, "Vote Buying and Accountability in Democratic Africa," PhD Dissertation at UCLA, 2013.

<sup>12</sup> It must be noted that in the pre-election survey, the sample was only divided into two random halves: a treatment and a control group. Overall, although there is some modification, the list-experiment that this study implemented was largely guided by Gonzalez-Ocantos, Ezequiel, Chad Kiewiet de Jonge, Carlos Meléndez, Javier Osorio and David W. Nickerson, "Vote Buying and Social Desirability Bias: Experimental Evidence from Nicaragua," *American Journal of Political Science*, Vol. 56 (1), No. 1, January, 2012: pp. 202–217. My experimental survey is also inspired by Kramon (2013) and De Jonge (2015).

13The possible presence of floor effects has long been central in the discussion among scholars. Blair and Imai (2012: 49–50) detect that floor effects may appear "if the control questions are so uncontroversial that uniformly negative responses are expected for many respondents." Another possible 'floor effect' may result if respondents fear that answering '0' reveals their honest (negative) preference (Ibid). There is some debate on this is, because on the other side of the coin, there may be 'ceiling effects' referring to a condition where all of the items would be acceptable to many participants. See Graeme Blair and Kosuke Imai (2012), "Statistical Analysis of List experiments," *Political Analysis* 20(1), 47–77.

Among those activities, how many had been carried out by candidates or brokers? Respondents could pick one, two, or three (all of them). I also provided an option for the scenario that 'none' of them had been carried out by candidates and their operatives. Meanwhile, each treatment group, which contained different sensitive issues, was presented a fourth different option, placed in the third response position.


Participants were then asked not to mention *which* items were true for them. They were only required to indicate *how many* of the items were true for them (see Kramon, 2013). Respondents could choose one, two, three, four, or all of them. Again, there was an optional answer for the scenario that 'none' of them had been carried out by candidates or their operatives. Following Gonzalez-Ocantos and his colleagues (2012) and De Jonge (2015), interviewees weren't asked to tell the field surveyor which particular activities they had experienced so that the surveyor wouldn't realise whether the respondent was reporting the sensitive item (receipt of cash or gifts) or not. The goal was to reduce social desirability bias (De Jonge, 2015).

Having previously shown the descriptive findings of the individual and neighbourhood measures, I now start the analysis and interpretation of the results from the list-experiment. The first column of Table 2.7 is the descriptive analysis to estimate people's experience of receiving handouts during campaigns. Table 2.7 shows the means for each experimental setting, namely one control group and three treatment groups. The list-experiment results found systematic differences in the means, providing a point estimate on the number of people reporting incidents of vote buying and coercion.


**Table 2.7** Descriptive analysis

The mean number of three campaign activities that include campaign posters or signs in the neighbourhood, party/candidate/broker visits, and campaign advertisements in television/newspaper/radio, reported by respondents in the control group, is 1.393. The average number of items indicated by the first treatment group, where subjects had the added option of "being coerced to vote for any political party or candidate," is 1.574. Subsequently, the mean number indicated by the second treatment group, where the subject had the choice of 'receiving money or gifts or favors,' is 1.668. Finally, the average number of items by the third treatment group which provided the additional option of a job offer for voters is only 1.416.

Thus, an estimation of the percentage of people reporting the receipt of electoral incentives can be established by comparing the average number of items indicated by the respondents in each group. Table 2.8 shows the treatment minus control differences. As the average number of items indicated by the control group is 1.393 and the average number of items indicated by the first treatment group is 1.574, we can conclude that 18.1% of respondents experienced coercion (1.574 − 1.393 = 0.181 and 0.181 × 100 = 18.1%). Random assignment assures that the difference is owing to respondents admitting coercion during elections. This number is clearly statistically significant (sig. < 0.05). Meanwhile, the mean number shown by the second treatment group (vote buying) is 1.668, producing a result of 1.668 − 1.393 = 0.274 and 0.274 × 100 = 27.4%.

Hence, importantly for this study, the estimation of the proportion of respondents receiving electoral incentives according to the list-experiment is 27.4%, much higher than the proportion of those experiencing coercion (18.1%). The difference is highly significant (sig. < 0.05). By contrast, clientelist exchange that took the form of a job offer to voters isn't statistically significant, with only a 2.2% difference between the control and



Source: This list-experiment was embedded within the post-election survey 22–26 April 2014

treatment groups. It can be concluded, therefore, that the practice of offering a job in exchange for a vote is rare in Indonesian electoral politics.

Interestingly, the estimated percentage of respondents reporting the receipt of handouts in this April 2014 list-experiment nearly tripled in comparison to a similar survey-based experiment carried out in March 2014, around one month before the election (see Table 2.9). Despite the sample being divided only into two random halves, the procedure of randomisation was exactly the same. In the pre-election survey, I also used a wording that was nearly identical to items employed in the post-election survey-based experiment. Table 2.9 reports the comparison between the results of the list-experiment in pre- and post-election surveys. As explained earlier, the estimated percentage of people receiving money or gifts according to the post-election survey-based experiment was 27.4%. Meanwhile, the mean number of campaign activities indicated by the control group with only four options in the pre-election experiment was 1.414, whereas the mean in the treatment group, where items had the additional option of picking the scenario of 'receiving money or gifts,' was 1.519. Hence, the intensity of vote buying that occurred about one month prior to the election was only 10.4% (Treatment − Control, 1.519 − 1.414 = 0.104, and 0.104 × 100 = 10.4%). The results from the pre- and post-election survey-based experiment are all statistically significant (sig. < 0.05).

The results thus far have found that that the estimates gathered from direct individual measures don't differ much from what is found in the list-experiment. The estimates derived from each method are strikingly similar, suggesting that in Indonesia, direct survey questions about individuals' experience receiving cash handouts aren't subject to response bias.


**Table 2.9** Vote buying in both pre- and post-election survey-based experiments

Source: The pre-election survey-based experiment was done during 19–24 March 2014, while the postelection experiment was conducted during 22–26 April 2014

My finding contrasts with that of Gonzalez-Ocantos and his colleagues (2012), who found substantial bias in a similar survey item asked in Nicaragua. My findings also contrast with those put forward by Corstange (2012) in his study of the 2009 Lebanese parliamentary elections and De Jonge (2015) in the case of the 2009 Honduran election. Both studies suggested that vote buying is found to be much more pervasive if detected via the list-experiment. My study, by contrast, confirms earlier work by Amick (2016: 1), who found that relative to a list-experiment, "direct survey questions to [Indonesian] voters about accepting transfers from campaigns elicits mostly honest responses from respondents."

Accordingly, my findings indicate that the difference between the direct individual and neighbourhood measures and the list-experiment isn't statistically significant. My list-experiment estimates about 27.4% of respondents actually sold their votes, compared to approximately 25% of respondents when asked directly about their personal experience of the 2014 elections. In every category of the individual-level questions discussed above, the estimates were statistically similar to those recorded by the list-experiment. Even the neighbourhood measure only gives a slightly higher estimate (28.9%) than the list-experiment.

Hence, the direct individual measure is reliable enough to estimate attempts of vote buying. This conclusion is supported by the fact that in the pre-election nationwide survey (March 2014), I found almost exactly the same estimates in the list-experiment and direct measure. In that survey, the list-experiment estimated that approximately 10.4% of respondents accepted electoral handouts, compared to 10.7% of respondents who answered in the affirmative through direct questioning. All of this suggests that analysis based on traditional obtrusive measures of vote buying is a valid and reliable approach in the Indonesian case. Both direct individual and neighbourhood measures are found not to understate the degree of such practices.

One plausible explanation for this lack of social bias is that Indonesian voters seem to be comparatively open about witnessing (and receiving offers of) vote buying. This openness, in turn, appears to be related to vote buying being less stigmatised than it was in the past (see Chap. 6). Despite its formal illegality, vote buying has increasingly become a normal transaction during election and is rarely prosecuted. In the eyes of many voters, an election is no longer viewed as a window of opportunity to express their political preferences, but rather as, borrowing Corstange's term (2012: 483), a 'season of money.' During my 13-month fieldwork at the height of the 2014 election, most politicians lamented the increased pressure on them to engage in vote buying, and they argued that such exchanges have become a part of routine politics in Indonesia. They admitted that voters suspected that whoever was elected—regardless from which party he or she originated—was highly likely to quickly forget their constituents after the election. Elections, therefore, are mainly seen as a 'temporary opportunity' for ordinary people to recapture material benefits that politicians have stolen (Kerkvliet, 1991: 231). In the words of Schaffer and Schedler (2007: 26), recipients view offers as "amends for [politicians'] wrongdoings [committed against them] in the past."

However, the initiative to 'normalise' vote buying doesn't only come from voters demanding handouts. To please such voters, and to avoid competing on programmatic grounds, many politicians happily present themselves as personal distributors of patronage, and they find moral justifications to defend this approach. For example, many Islamic politicians justify doing so by quoting religious scholars (*kiai*). For instance, one successful local candidate in Central Java relied on the advice given by an influential *kiai*. This *kiai* stated that if he didn't distribute cash and lost as a result, all his other efforts to win office would be proven meaningless. The *kiai* reportedly quoted an Arabic script, taken from one of the major textbooks (*kitab*) studied in Islamic boarding schools, that in his view allowed vote buying. The passage reads: (Bribing voters is basically unlawful, but it could be permitted for fairness). Additionally, the *kiai*—who sat on the advisory board of the candidate's moderate Islamic party—cited an Islamic jurisprudential premise, namely (that without which an obligation cannot be fulfilled is itself obligatory). Based on this proposition, the ultimate goal is the election of a good candidate motivated by a religious cause. If good candidates are reluctant to spend money for vote buying—so the argument goes—and lose as a result, then parliament will be made up only of corrupt politicians (Interview, 12 August 2014). In this view, vote buying is simply an electoral strategy for winning and thus preventing corrupt politicians from taking office. Accordingly, it isn't only justifiable but also necessary.

As a consequence of such efforts at 'normalisation' of vote buying, the stigma attached to it has weakened. During my research, most people talked about vote-buying practices in an open way, indicating that there is little social desirability bias. As Agun Gunandjar Sudarsa of Golkar put it, "vote buying has become public knowledge in Indonesia, [and is] seen as part of a tradition during elections. Indeed, it is illegal on paper, but I am sure the police will not pursue it. Otherwise, the prisons would be full" (Interview, 23 April 2014). Terms such as 'NPWP' (*Nomer Piro, Wani Piro*, representing the question 'what number on the ballot are you and how much do you dare to give?') or GOLPUT (*Golongan Penerima Uang Tunai*, roughly a 'group of cash recipients')14 circulated widely, pointing to the centrality of such clientelist exchanges. The famous but unsuccessful incumbent candidate from Golkar, Nurul Arifin, stated that in a context where the scale of vote buying is so extensive, the only way to beat co-partisans is by outspending one's internal party rival in terms of distributing cash to voters (Interview, 28 April 2014). Thus, vote buying appeared, borrowing the words of an independent researcher, to be "more massive, vulgar and brutal in the 2014 elections" than ever before (*Kompas.com*, 21 April 2014).

Far from being increasingly socially stigmatised, then, vote buying in Indonesia is deeply ingrained in the society's fabric. As a result, as argued by Benedict Kerkvliet (1991: 80), for the Philippines, voters see elections as a strategic transaction with power, showing that they too "as voters aren't merely subjects of the seemingly powerful candidates or politicians."

#### 2.6 Vote Buying in Local Executive Elections

To further gauge the extent and social acceptability of vote buying in Indonesia, and to compare how such practices play out in different settings, it is important to assess whether different datasets are *consistent with* the primary *data source* used in this study. As noted earlier, the previous sections relied heavily on the 2014 post-election, nationwide survey. In the remainder of this chapter, I present a large amount of data drawn from 963 local surveys conducted by my home institutions LSI and Indikator as well as by SMRC between 2006 and 2015. Over the entire period, these three preeminent polling organisations conducted surveys about regional executive elections (*pilkada*) in 34 provinces and 513 regencies/cities across Indonesia. Though not their primary focus, these surveys included common questions related to vote buying. The relevant question was: "As an effort to win the *gubernatorial/regency/mayoral* election, certain candidates or brokers typically give money or gifts for people to influence their vote. In your opinion, can the money/gift be considered acceptable or unacceptable?"15 This wording is unobtrusive in that it doesn't inquire about whether vote buying took place or not. But, at least, it may serve to proxy the extent to which vote buying is viewed as acceptable in local executive elections.

<sup>14</sup> In normal usage, NPWP is the acronym for 'tax file number,' while GOLPUT denotes those citizens who deliberately abstain from voting.

15The wording of the question wasn't always the same. The categories in italics changed according to the level of the territory in which the election took place (i.e. in a province, city, or district).

My focus in analysing this dataset is on inter-regional variation in subnational elections, which I couldn't capture in the national surveys discussed thus far. Note that the population sizes at the district and provincial level are extremely varied, and the frequencies of the surveys were uneven, which could bias the sample to certain regencies/cities. In order to correct this non-random geographic sampling, the weighting scheme is carried out in two stages. In the first stage, the weighting is applied proportionally at the provincial level. In the second stage, the weighting is made within each province to adjust for the over- or under-sampling of voters from certain regencies/cities within provinces. By doing so, the sample is representative at the regency/city level and enables me to generate district-level estimates. Put simply, the formula used to weight observations is:

$$W\_{\vec{g}} = \pi\_{\vec{\wedge}} \wedge p\_{\vec{\wedge}}$$

In which *πj* = proportion of population in each regent/city *j*

*pj* = proportion of sample in each regency/city *j*

Table 2.10 shows the degree to which vote buying was reported as acceptable by respondents in these local executive election surveys through the period 2006–2015. The acceptability of vote buying at the local level is relatively high, with, overall, four out of ten Indonesians finding it acceptable for politicians or their brokers to distribute cash or gifts as part of their campaigning. Although 60.2% thought vote buying was unacceptable, the average percentage of respondents who said such practices was acceptable is high. Using similar data from Timor-Leste, for instance, 'only' 32.7% of the electorate in the country thought that vote buying was acceptable.16


**Table 2.10** Reported acceptance of vote buying, 2006–2015

16Lembaga Survei Timor-Leste (The Timor-Leste Survey Institute) and Lembaga Survei Indonesia (LSI), "Exit Poll: Eleisaun Parlamentar Timor-Leste," 22 July 2017.

My study found substantial differences between regions. Those who lived in Java were more likely to consider voting buying to be acceptable. The data show that 46.3% of respondents from Java didn't have a problem accepting cash or gifts from would-be regional heads. This compared to 'only' 31% of people from the outer islands who thought vote buying was acceptable. Further, Table 2.11 hints at some intriguing heterogeneity across regions. Relative to other outer island regions, people who lived in Sumatra were less accepting of vote buying in local executive elections.

There are interesting variations across provinces as well. The average percentage of Sumatrans who thought vote buying was acceptable is 28.1%. Some provinces scored significantly above the average, such as Lampung (40.4%), Bengkulu (38.6%), and South Sumatra (32.3%). Meanwhile, some provinces exhibited well below the regional average, such as West Sumatra (13.1%) and Riau Islands (16.4%). Given that the data are representative at the district level, we can assess further which regency or city in each province has the highest levels of vote-buying acceptability. In Aceh, for example, despite the province's overall exhibiting acceptance levels well below the Sumatran average, the reported tolerance of vote buying was greatest in Pidie Jaya (70%), Aceh Singkil (50%), and East Aceh (36.7%).

There is also a wide variation in the degree of reported acceptance of vote buying in Java. Among all provinces in Java, voters who lived in Central Java were most likely to be tolerant of vote buying, with 50.9% of respondents considering vote buying as a normal practice during local elections. Following Central Java are Banten with 46.6% and East Java with 46.5%. At the other end of the spectrum, Jakarta is among those provinces in Java in which voters were less accepting of vote buying, with only 24.7% answering in the affirmative. Again, substantial heterogeneity was found within each province. Despite its notoriety as the province that is most tolerant of vote buying, Central Java wasn't homogeneous. Voters


**Table 2.11** Reported acceptance of vote buying by region (%)

Source: LSI, Indikator, and SMRC historical dataset on sub-national executive elections

who lived in cities (*kota*) such as Surakarta, Semarang, and Magelang, for instance, were less accepting of vote buying than those in regencies (*kabupaten*) in that province.

Further, in some cases, my datasets include information on the acceptable price of the votes. Among those who thought vote buying was acceptable, I asked a follow-up question: "How much cash would a candidate need to give for you to find it appropriate to vote for him/her?" As Table 2.12 shows, while responses were quite scattered, the most frequently given answer was 'less than IDR 50,000' (approx. US\$4.40). Note that those who 'don't know' or 'refuse to answer' aren't included in the analysis.

Interestingly, expectations about the appropriate price of a vote in Java were much lower than in the outer islands. For instance, only 12.9% of respondents who live outside Java17 saw it as appropriate to vote for a candidate if given less than IDR 50,000, while a significant proportion of Javanese reported that small sums of money would be satisfactory. Only 9.4% of respondents in Kalimantan would be happy if given less than IDR 50,000 (Table 2.13).

As I will argue in Chap. 5, the smaller amount that a candidate must spend on 'buying' votes in Java should be largely read in terms of its specific moral economy of gift giving, which is centred around the concept of *sangu*. In Javanese, *sangu* means pocket money or food that is usually taken on a journey as supplies. Applied to the context of vote buying, it isn't necessarily the size of the payments that counts, but rather its symbolic appropriateness. As a one-time DPR member who ran in an electoral


**Table 2.12** The price of a vote (%)

Source: LSI, Indikator, and SMRC historical dataset on sub-national executive elections

17This is the number of the entirety of the outer islands calculated by adding together the rates of those who live in four regions (Sumatra, Kalimantan, Sulawesi, and others) thought the acceptable price of the votes was less than IDR 50,000 and then dividing this total by four.


**Table 2.13** The price of votes by region (%)

Source: LSI, Indikator, and SMRC historical dataset on sub-national executive elections

district in West Java put it: "If you are seeking people's votes, you have to have good understandings. People who are going to vote for me would be unable to work on voting day. I compensate them for that. It isn't expensive, just souvenirs to 'tie' them so that they don't run to cashed-up rivals" (Interview, 24 April 2014).

In addition to the cultural context, the price of vote buying is greater in local elections where the constituency size is smaller. This helps explain why average expectations of the price of votes in the outer islands were higher because constituencies there are smaller than on densely populated Java. The smaller the population of the district or city, the more a candidate has to spend to buy a vote. The price escalates in some resource-rich, but sparsely populated districts on Kalimantan or in the eastern part of Indonesia. For instance, my study found that 82.1% of voters in Tana Tidung (North Kalimantan) would consider more than IDR 200,000 (approx. US\$17.80) an appropriate price for a vote. This is one of Indonesia's most resource-rich districts, with significant production of oil, natural gas, and coal, but its population is only about 14,899, escalating the vote-buying rate for a single candidate to at least IDR 1,000,000 (approx. US\$89) (Informal Communication with a national party leader, 21 May 2014).

Put in a nutshell, the market for votes is huge in Indonesia and vote buying has been pervasive not only in national legislative elections but also in local executive elections. One broker who worked for a winning candidate during the 2017 direct election for regency head in Batang, Central Java, acknowledged that distributing cash handouts was part of his chief strategy because voters saw elections as a 'money harvest' (Interview, 23 January 2017). Prior to the first simultaneous local executive elections in December 2015, I interviewed a seasoned broker in East Java. He revealed two important indicators of the intensity of vote buying. First, because cash handouts are usually distributed in plain white envelopes, along with the name card of the candidate, the sales of envelopes in election times increase sharply.18 For example, based on the information he received, the sale of envelopes in Sumenep during the 2014 legislative elections rose by 75%, in Sidoarjo by 65%, in Gresik by 60%, Mojokerto by 80%, Ponorogo by 70%, and Trenggalek by 80%. A second anecdotal indication of the prevalence of vote buying is a shortage of small banknotes as election day nears. According to the above broker, most candidates want to change money into smaller denomination banknotes at the bank because vote buying typically takes the form of small payments (usually around IDR 10,000–20,000) (Interview, 5 December 2015).

Overall, as this section has shown, vote buying has been an integral part of Indonesian electoral politics, not only at the national level but also at the local level. Complementing the 2014 national voter survey discussed in previous sections, the sub-national election dataset confirms that vote buying has played a key role in mobilising electoral support. Using the massive dataset of local election surveys that covers the period 2006–2015, I have shown that the level of reported acceptance of vote buying is high, with four out of ten Indonesians not having a problem accepting cash or a gift from candidates for regional government heads. All of this suggests that vote buying has become one of the main instruments of electoral mobilisation in local government contests.

### 2.7 Conclusion

This chapter has explored the prevalence of vote buying in Indonesian election campaigns. Elaborated from a wide range of methods—whether individual, observational, or derived from the list-experiment—the results are mostly congruent with each other. The findings generated from the direct individual and neighbourhood measures as well as from the survey-based experiment are that 25% of Indonesian voters were exposed to vote buying in the 2014 legislative election when asked directly, 27.4% when asked via the list-experiment, and 28.9% when asked through the neighbourhood question. If we use the less obtrusive

<sup>18</sup>This observation is based on information he collected from big shops in each regency that sell envelopes on a large scale.

measure without concretely mentioning the 2014 election (as discussed in detail above), the level of vote buying rises slightly to 33%. Hence, the estimated proportion of people engaging in vote buying in Indonesia lies between 25% and 33%, considering that the difference between one measure to another isn't statistically distinguishable.

In the 2014 legislative election, there were around 187 million registered domestic voters. A range of between 25% and 33% would mean an estimated 47 million to 62 million voters nationwide were offered cash or other material benefits in return for their votes. If we rely on the highest estimate, one out of three Indonesian voters was personally exposed to vote buying. It is noteworthy, however, that these figures define a range, rather than a precise point estimate, of vote-buying incidents. In addition, vote buying wasn't only a prominent feature of Indonesia's national legislative elections. My large dataset of local elections across the country found fewer than four out of ten Indonesians thought of vote buying to be an acceptable practice. However, my study found empirical evidence that such practice in general is ubiquitous in Indonesian electoral politics.

Such findings generate more questions. Most obviously, given that vote buying is common but not universal: which voters are targeted in votebuying exchanges? The next chapter provides a comprehensive profile of the typical vote sellers and presents more rigorous analysis to predict the likelihood of a person being offered benefits in exchange for a vote. It does so by testing large number of variables that are generally believed to be the determinants of vote buying.

#### Bibliography

Books, Journal Articles, and Unpublished Papers


Articles and Reports

KPU. (2014) Partisipasi Pemilih pada Pemilu 2014. Viewed at http://www.kpu. go.id/koleksigambar/Partisipasi\_Pemilih\_ pada\_ Pemilu2014\_ Studi\_ Penjajakan.pdf. Accessed 1 November 2014.

**Open Access** This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

# The Determinants of Vote Buying: The Profile of Typical Vote 'Sellers'

In the preceding chapter, I showed how prevalent vote buying is in Indonesian elections, even by international standards. The findings offered a systematic confirmation of thus far largely anecdotal accounts of political clientelism in Indonesia. Despite the ubiquity of such practice and a recent surge in publications on the topic (e.g. Aspinall, 2014; Allen, 2015; Aspinall and Sukmajati, 2016; Amick, 2016; Aspinall et al., 2017; Tawakkal et al., 2017), surprisingly the question of what types of individuals are targeted by vote buying remains largely unexplored. Accordingly, this chapter focuses on the analysis of the individual-level determinants of targeting of vote buying and provides a comprehensive profile of the typical vote 'sellers.' I examine a wide range of variables that are generally believed to be the determinants of such electoral strategies. The final part of this chapter identifies the most striking aspects of vote buying in the national legislative elections.

One of the key findings of this chapter is that voter identification with political parties is consistently and significantly linked to vote buying. The closer the ties of a voter to a political party, the more likely that voter is to receive offers of vote buying. My post-legislative election data also shows that, contrary to prior expectations, most of the variables associated with modernisation theory have little correlation with vote buying. The insignificance of socio-economic factors indicates that the targeting of vote buying in the national parliamentary elections can be best explained in terms of party identification rather than the modernisation paradigm.

### 3.1 Perspectives on Electoral Clientelism

In comparative studies on electoral clientelism, discussion of the determinants of vote buying at the individual level typically focus on the issue of target selection. As discussed in Chap. 1, because they have limited budgets, political machines aim to efficiently spend their resources on those voters most likely to respond positively to vote-buying attempts (i.e. voters who are most likely to commit their vote as a result of a vote-buying operation). Additionally, since ballot secrecy prevents candidates and brokers from unequivocally verifying whether recipients of payments reciprocate with their votes, it is essential for them to identify those voters who aren't only most likely to be influenced by vote buying but are also most *reliable* in delivering a vote.

Broadly, the literature on electoral clientelism has identified three factors explaining why some individuals are more likely than others to be targeted with vote buying: socio-economic and demographic factors; levels of civic engagement; and citizens' political attitudes. The first school of thought focuses on a quasi-determinist view of electoral clientelism that is congruent with arguments made by modernisation theorists. According to this camp, clientelism is best described as a pre-modern form of political and social relations, involving mostly lower class citizens (e.g. Lipset, 1959; Scott, 1972). Thus, it is an intrinsic element of 'third world' politics, affecting countries that are relatively poor and have low rates of literacy. They aren't 'modern,' say Andrews and Inman (2009: 6). Flowing from this conception of clientelism is the inverse argument that successful democratisation is only possible in polities with sufficiently high levels of economic development. Accordingly, as Hicken (2011: 299) explains, "clientelism appears to be more prevalent in developing countries, and within nations, noting that poorer voters appear to be more susceptible to clientelist offers than richer voters."

Many scholars therefore draw a connection between different dimensions of socio-economic modernisation and the prevalence of clientelist exchanges (e.g. Jensen and Justesen, 2014; Brusco et al., 2004). Such scholars believe that poor people are more attracted by vote buying than wealthier citizens. Stokes (2007b: 618) introduces the model of a high discount rate, explaining that "poor people are risk-averse and hence value more highly a bag of goodies in hand today than the promise of redistributive public policy tomorrow." The poor discount future programmatic benefits because future rewards are less certain and concrete than the ones they can collect now (Stokes, 2007a: 94). The literature on electoral clientelism has also emphasised the effects of education, proposing that low education makes citizens more vulnerable to vote buying (Kitschelt and Wilkinson, 2007).

Still within this modernisation paradigm, some scholars have suggested that variation in patterns of vote buying can result from different places of residence. For example, Hicken (2007a: 56) argues that vote buying is less likely in urban areas, partly due to the diminishing role in such settings of traditional patron-client networks through which a candidate can deliver material benefits. In addition, according to Hicken (2007a: 56), people who live in rural areas are more likely to be prone to vote buying because income and education levels are higher in urban areas and there are greater demands for public goods in urban areas, where residents are concerned about issues such as traffic congestion, public transportation, and garbage collection.

The focus of many authors advancing the economic modernisation approach has been on the conceptualisation of the argument and case studies that support it (Hicken, 2011; Scott, 1972; Jensen and Justesen, 2014). Based on these studies, as Hicken (2011: 297) put it, "clientelism was bound to disappear as countries modernised both economically and democratically." For that reason, vote buying appears to be endemic in many developing countries—and almost exclusively limited to them too. Although the assumptions put forward by the modernisation camp seem to be plausible, the developmentalist argument fails to explain why clientelist networks remain influential in relatively wealthy and highly educated nations, such as Japan (Kitschelt, 2007), Belgium (Kitschelt, 2007), Austria (Kitschelt, 2007), and Italy (Kitschelt, 2007). Such cases challenge the conclusions of the modernisation school within clientelism studies.

A second group trying to explain which voters get targeted most by vote buying has concentrated on civic engagement. In order to facilitate clientelistic exchange, many politicians make use of informal and formal organisations with large numbers of members. As a result, people who are involved in mass organisations are expected to be particularly exposed to vote buying. In the Philippines, Cruz (2014) argues that voters who are actively engaged in social networks are disproportionately targeted for vote buying. Gonzalez-Ocantos and his colleagues (2012: 212) found a similar conclusion, showing that Nicaraguans who actively engage in meetings held by civic associations are "far more likely to report vote buying than those who never attend." This is partly because such social organisations provide norms of reciprocity that can be used to mobilise support through particularistic electoral mobilisation and to ensure that recipients of patronage actually reciprocate with their votes (Callahan, 2005: 496).

Similarly, in the case of Thailand, Callahan (2005) suggests that civic associations are responsible for the emergence of new forms of electoral corruption and vote-buying practices at election times. As theoritised by Putnam (1993: 167), networks of civic engagement and norms of reciprocity are highly correlated since the two are an important feature of social capital. In this concept, norms of reciprocity are one vital aspect of the social capital that can facilitate clientelist exchanges. Cultural norms of gift giving, according to this approach, create a social atmosphere conducive to quid pro quo offers. This atmosphere, in turn, breeds a culture of mutual favours in which parties or candidates are happy to buy votes and voters are happy to 'sell' their vote in exchange for payment or other material benefits. Finan and Schechter (2012), therefore, argue that vote buying is inherently self-enforcing due to the existence of social norms of reciprocity.

A third camp of scholars has emphasised a potential link between electoral clientelism and political attitudes (e.g. Manzetti and Wilson, 2009; Banegas, 1998; Carreras and Irepoglu, 2013). Generally, political attitudes include components such as party identification, efficacy, political interest, political information, political trust, political participation, and support for democracy (Verba et al., 1995). Regarding party identification, scholars have offered no definitive conclusion on whether this component of political attitudes predicts the likelihood of receiving offers of vote buying. On the contrary, this issue is hotly debated. As we have seen, some scholars argue that payment for votes targets swing, or ideologically indifferent, voters (meaning that strong party identification would be negatively correlated with vote buying). As discussed thoroughly in Chap. 1, some even argue that providing money or private goods to those who are close to, or ideologically proximate to, any political party is wasteful (Lindbeck and Weibull, 1987; Dixit and Londregan, 1996; Stokes, 2005). In contrast, other authors suggest that political parties tend to focus their vote-buying efforts on their partisan, loyal voters whose turnout can be maximised—which would mean that high levels of party identification should predict vote buying (Cox and McCubbins, 1986; Nichter, 2008; Gans-Morse et al., 2014; Diaz-Cayeros et al., 2012; Stokes et al., 2013).

Another dimension of political attitudes that might be relevant to political clientelism is political efficacy, that is, the citizenry's ability to influence government decisions, their belief that government cares what people think, and belief that government will respond to their demands. Reef and Knoke (1999: 414) define political efficacy as "an individual's sense of personal competence in influencing the political system." Political efficacy is adversely related to political alienation or political powerlessness, which refers to "a person's perceived inability to influence governmental policy" (Ibid., 414). Some scholars found a link between low levels of efficacy and vote buying. In Benin, for instance, material rewards offered by parties during elections are seen as an expression of political alienation, in which vote sellers were more likely to feel powerless about government and perceive politics as meaningless (Banegas, 1998: 78–79). Kerkvliet (1991: 231) found a similar pattern in the Philippines, where villagers tend to receive material benefits as "practically their only opportunity to get anything from people in government." In this view, disempowered citizens view elections as a momentary opportunity to "stake a rightful claim to the resources of those higher up" (Schaffer and Schedler, 2007: 26).

A further political attitude dimension that potentially correlates with electoral clientelism is political information. Grossman and Helpman (1996) suggest that 'uninformed voters' will be the most likely targets of clientelist campaign strategies. Vicente and Wantchekon (2009: 302) come to a similar finding, arguing that informed voters are more likely to support politicians who run programmatic campaigns and will stay away from material, particularistic strategies.

The next potentially relevant component of political attitudes is political interest, defined as "the degree to which politics arouses a citizen's curiosity" (Van Deth, 1989: 278). Some scholars suggest that linkages may exist between such interest and exposure to clientelist exchanges (e.g. Carreras and Irepoglu, 2013). An interested citizen, according to this view, is likely motivated to participate in elections without material rewards. This psychological variable is therefore believed to have a negative relationship with clientelist mobilisation. Moreover, an interested citizen is assumed to be efficacious and informed, linking political interest with other factors mentioned above. Generally, interested voters are expected to support parties or candidates who rely on public policy campaigns, while uninterested voters are expected to respond to vote-buying exchanges.

Another related component of political attitudes that has started to receive scholarly interest in clientelism is political trust (Manzetti and Wilson, 2009). Political trust is defined in the literature as a basic evaluative orientation towards the political system (Citrin and Muste, 1999; Inglehart, 1999). While related to political efficacy, trust levels are more generally about the citizenry's evaluation of existing political institutions. In short, political efficacy is an input, while trust in institutions is an output of the political system (Almond and Verba, 1963). Manzetti and Wilson (2009) argue that clientelist politics and trust in political institutions are strongly related. They found that in countries where political and government institutions are weak and patron-client relationships are strong, voters tend to support corrupt leaders or parties from whom they expect to receive material benefits. Accordingly, voters with lower levels of trust towards political institutions are believed to build transactional relations with political parties and 'sell' their votes to candidates providing private goods and favours.

Electoral participation is another political attitude dimension that has become a matter of debate in the literature on clientelism. There has been a growing discussion about whether vote buying increases or decreases voter participation in elections. Using a field experiment in West Africa, Vicente (2013) found strong evidence that vote buying increases participation. Similarly, Carreras and Irepoglu (2013: 616) suggest that the distribution of electoral rewards is effective in mobilising voters in Latin America. In Egypt, Blaydes (2011) concludes that voters exercised their right to vote because they expected benefits (to be sure, she was describing elections under Mubarak's authoritarian rule). Nonetheless, other studies have produced contradictory findings. In Nigeria, Bratton (2008: 15), for instance, found that vote buying decreases individuals' electoral participation, suggesting that along with electoral violence, clientelist exchanges trigger disillusionment among the electorate, leading them to exit the political process.

A final attitudinal variable that might interrelate with political clientelism is support for democracy. It is generally held among scholars that clientelist exchanges are inimical to democracy (Stokes, 2005; Keefer, 2007; Kitschelt and Wilkinson, 2007). Stokes (2005: 316) argues that vote buying is undemocratic because it involves 'perverse accountability.' Instead of politicians being accountable to voters, she explains, where vote buying happens, voters are held accountable for their vote. Hence, it might be inferred that those who strongly support democracy would be less likely to engage in vote buying, and vice versa.

In sum, there are three broad conceptual frameworks that might affect the magnitude of vote buying: socio-economic factors, involvement in civic organisations, and political attitudes. In the next sections, I draw specific hypotheses from these three broad areas of inquiry and test them by analysing the dataset available to this study.

#### 3.2 Hypotheses

#### *3.2.1 Vote Buying and Modernisation Theory*

As noted above, many scholars believe that socio-economic factors, such as income level and education, shape how widespread vote buying (Brusco et al., 2004; Vicente, 2013) becomes. Poor voters are believed to be significantly more vulnerable to such practices than wealthier ones. If this is true, then I expect to find that *the poorer a voter, the more likely he or she is to experience vote buying*. Another important aspect of modernisation that is believed to correlate with exchanges of material benefits for votes is education (Vicente, 2013). Indeed, variables of education and income are often highly correlated: people with lower education usually generate lower incomes than those with higher education, and vice versa. Consequently, less-educated individuals are expected to be more intensively exposed to vote buying than well-educated people. The hypothesis is: *the less educated a voter, the more likely he or she is to experience vote buying*. Additionally, as we have seen, differences in voters' geographical location are also believed to influence the potential for electoral clientelism. Here, the hypothesis is: *a voter who lives in a rural area is more likely to experience vote buying than a voter residing in an urban area*.

#### *3.2.2 Vote Buying and Civic Engagement*

As explained above, networks of civic engagement are typically used by clientelist actors to distribute patronage resources (Callahan, 2005). Virtually everywhere in the world, vote buying isn't legal. Therefore, politicians pursuing vote buying are forced to be discreet. However, it is difficult to engage large numbers of voters discreetly with vote-buying attempts (Hicken, 2007a, 2007b). Accordingly, they find ways to disguise their vote-buying attempts by, for instance, penetrating informal and formal organisations and packaging their clientelist offers in religious or social assistance terms. Such patterns have led to claims that social organisations—and their members—are more likely to be targeted by vote buying than individual voters without such links. If this is correct, *the more deeply involved a voter is in social organisations, the more likely he or she is to be offered vote buying*.

#### *3.2.3 Vote Buying and Political Attitudes*

We have also seen that extant explanations for electoral clientelism also include political attitudes, comprising several components. Concerning party identification, some scholars believe that voters who don't identify with a particular party (i.e. swing voters) tend to receive more material rewards. If this proposition is true, I expect to find that *the more nonpartisan a voter is, the more likely he or she is to accept vote buying*. A negative finding with regard to this hypothesis would direct us towards the core-voter model. Another aspect of political attitudes is political efficacy. With respect to efficacy, much of the literature suggests that clientelist exchanges are an indication of lower efficacy or an expression of political alienation. The hypothesis for this study, then, is *the lower efficacy a citizen has the more likely he or she is to engage in vote buying*.

Similarly, some literature claims that uninformed voters are more likely to be targeted by vote buying. If this claim is correct, we must hypothesise that *individuals with increased access to information are less likely to receive offers of vote buying*. The literature also suggests that voters with low levels of political interest are more susceptible to clientelist offers. If so, *the more uninterested a citizen is in politics, the more likely he or she will be experiencing vote-buying attempts*. With regard to political trust, there is an increasing perception that low levels of trust in political institutions lead to higher levels of vote buying. If so, it is plausible to hypothesise that *the lower a voter's trust in political institutions, the more vulnerable she or he is to vote buying*. In addition, much of the literature on clientelism also claims that participation in elections increases the likelihood of being targeted clientelist offers. If so, *voters who participate in the election are more likely to experience vote buying*. Finally, a widely held view among scholars is that clientelism runs counter to democratic principles, implying that those who strongly support democracy will stay away from such exchanges. Thus, I expect to find that *those who support democracy are less likely to engage in vote-buying operations*.

## 3.3 Measures of Modernisation Theory, Civic Engagement, and Political Attitudes

In order to test the hypotheses developed above, this section introduces the independent variables needed to calculate the validity of the claims. As presented in Chap. 2, the dependent variables of my post-election survey in April 2014 are an *experience*, that is, being targeted by vote buying used to determine factors that explain why some individuals were more likely than others to receive offers of vote buying. The central independent variables, on the other hand, include socio-economic demographics, civic engagement, and a set of political attitudes, including party identification.

#### *3.3.1 Modernisation Theory*

I employ three common indicators of the modernisation argument: poverty, urban-rural domicile, and education. First, in the dataset, I use a single 12-point scale item to measure an individuals' income level. This measure asked: "On monthly average, how much is the gross income of your household?" Income is a continuous variable that reports the respondent's gross household income per month, coded on a scale from 1 to 12, where 1 indicates that the respondent's income per month is under IDR 200,000.00 (approximately US\$17.60) and 12 indicates over IDR 4,000,000.00 (approximately US\$352).

Second, it also identifies respondents' domicile based on a simple ruralurban dichotomy. This dichotomy, in turn, is drawn from an official ruralurban category developed by Indonesia's Central Agency of Statistics (BPS, Badan Pusat Statistik). I measured rural-urban residence by constructing a dummy variable that indicates whether respondents live in a rural area (coded 1) or in an urban area (coded 0). Examining income level and urban/rural residence simultaneously is particularly important in Indonesia, since poverty is predominantly concentrated in rural areas; in the early 2010s, 16.6% of rural people were poor, compared with 9.9% of the urban population (Rural Poverty Portal, 2012). In my sample, which was based on the last census data from the government, 50.2% of respondents lived in rural areas, while 49.6% lived in urban areas.

Third, my data source employs education as the final dimension of modernisation, asking respondents about the highest level of education they have completed. Education is coded 1–10, where 1 indicates that the respondent never attended school and 10 indicates that the respondent held a bachelor's or higher degree. A plurality of respondents (41.4%) reported that they only had primary education, hadn't completed formal education, or had no formal education at all, while 21.1% reported finishing junior high school, 26.3% reported finishing senior high school, and 11.1% reported having higher education.

In addition, the dataset includes other demographic variables such as gender, age, ethnicity, and religion. First, it controls for the respondents' gender to capture any possible relationship between vote buying and gender. Gender is coded 1 if the respondent is a male and 0 if the respondent is a female. As described in Appendix A, the sample was constructed to include 50% of males and 50% of females, mirroring the last census data in 2010. Second, the dataset controls the age of respondents to identify any bias in the tendency of vote buyers to particularly target young (or older) voters.1 Age is a continuous variable that reports the respondent's age in years, ranging from the youngest to the oldest. Respondents who were aged 21 years old or less constituted only 4.3% of the sample, while 44.7% reported that ages ranging between 22 and 40 years. The percentage of those aged 41 to 55 years accounted for 34.7%, and 16.3% were older than 55 years.

Another potentially relevant factor predicting vote buying is ethnic identity. Because the Javanese are the largest ethnic group in Indonesia (making up approximately 40% of the total population), ethnicity is therefore divided into two categories only: Javanese and others. Selfidentification as Javanese was coded as 1, while others were coded 0. About 41% of respondents reported being of Javanese ethnicity. The rest of the respondents identified themselves as non-Javanese; rather, they were affiliated with a large number of small ethnic groups across the archipelago. The last demographic variable is religion. Since other religions are small relative to Islam,2 religious affiliation is divided into two categories

<sup>1</sup> Some scholars found evidence of a link between age and voters' susceptibility to vote buying (e.g. Brusco et al., 2004). In Argentina, the younger cohorts among low-income Peronists are more likely to receive vote-buying offers than older cohorts. The rationale is that these differences could reflect life-cycle effects, whereby older voters are more likely to be more partisan and have higher party identification levels than younger ones. Their findings support the swing-voter argument, suggesting that political machines tend to target ideologically indifferent voters because they are more dependent on material rewards.

2Around 202.9 million people identify themselves as Muslim, which represents 87.3% of Indonesia's total population today. In my sample, 89.1% of respondents reported that they were Muslim, 8.4% were Christian/Catholic, and 2.4% others. See Appendix A.

only: Islam and others. I constructed a dummy variable where respondents who identify as Muslim were coded 1 and 0 if from a different religion. In the legislative dataset, given that the sample drawn from the majority of provinces in outer islands was small relative to provinces located in Java, the regional variable is divided into two categories: Java and others.

#### *3.3.2 Civic Engagement*

Putnam and Goss (2002: 10) postulated that networks of civic engagement comprise informal as well as formal organisations. They suggested that such networks don't only take the forms of civic associations but also manifest in informal social engagements such as dinners with friends, gathering in a café, and so forth. Following Mujani (2003), this study examines the degree of civic involvement by establishing the level of respondent's engagement in religious and non-religious organisations. The prompt was the following question: "Allow me to inquire about your participation in any of the following organisations or groups below. Are you an active member, inactive member or non-member of these organisations?"3 The available choices include religious organisations, such as Nahdlatul Ulama (NU) and Muhammadiyah in the context of Islam or churches in the case of Christianity, or 'secular' associations, such as youth organisations and sport clubs; agricultural and fishermen's groups; and labour unions, political parties, youth community councils (*karang taruna*), regular social gatherings, cooperatives, art and cultural clubs, and others.

#### *3.3.3 Political Attitudes*

In the dataset, measures of political attitudes include party identification, efficacy, political interest, political information, political trust, political participation, and support for democracy. In the following sub-sections, I describe in detail the measures developed for each component in the broader area of political attitudes.

3For coding and scaling purpose, each item of civic engagement comprises a three-point scale: non-member (0), non-active member (0.5), and active member (1). Scores for membership in the association are then added up and divided by 12 to create a three-point scale of networks of civic engagement.

#### *3.3.3.1 Party Identification*

My post-parliamentary election dataset measures the degree of party identification by using two items. The first question regarding party attachment was: "There are people who feel closer to a certain political party and some who aren't. How about you, do you feel there are any political parties who you feel closer to?" Those who responded affirmatively were asked to name the specific party they feel close to. They were also requested to rate the strength of affiliation to their party on a three-point scale: "How close do you feel toward the party?" In the analysis, overall party identification is an additive scale from these items which is then normalised to a scale of between 1 and 4, in which 1 indicates respondents who don't feel close at all to any party, 2 indicates those who feel somewhat close, 3 reflects those who feel quite close, and 4 reflects those who feel very close.

#### *3.3.3.2 Political Efficacy*

The literature on political behaviour divides political efficacy into two forms: internal and external. The former refers to a citizen's belief that he or she is able to understand politics and influence government decisions, while the latter deals with how a person feels the government will respond to his/her demands (Reef and Knoke, 1999: 414). For internal efficacy, I used responses to the items: (1) "People like me cannot influence decisions by government" and (2) "In general, political issues are too complex, so people like me cannot understand what is going on."4 For external efficacy, I used a single item, asking whether it is true that "people like you aren't heard by political leaders." Each of these items is a four-point scale, with responses ranging from 'strongly disagree' to 'strongly agree.' Since all three questions were negatively keyed, I recoded them with reversed Likert scales.5 By reverse-scoring all of the negatively keyed political

4For further discussion, see Niemi, Craig, and Mattei on "Measuring Internal Political Efficacy in the 1988 National Election Study," *The American Political Science Review* Vol. 85, 4 (December 1991), pp. 1407–1413.

<sup>5</sup> In statistical analysis, positively keyed items are understood as items that are worded so that an agreement with the item represents a relatively high level of the measured attribute. In contrast, negatively keyed items are defined as those that are worded so that an agreement with the item represents a relatively low level of the measured attribute. The objective of reverse-scoring is to make sure that all the items are consistent with each other—both the originally negatively keyed items and positively keyed items—with regard to the implications of an 'agree' or 'disagree' response. For discussion, see R. Michael Furr's note at http:// psych.wfu.edu/furr/716/Reverse-scoring.doc.

efficacy items, I created consistency among the items. Adding the score of all three items and dividing the result by three produced a four-point scale of political efficacy, which ranged from not efficacious at all (1) to very efficacious (4).

#### *3.3.3.3 Political Information*

Political information is measured by responses to five questions. The first question was "How often do you follow the news on politics or government?" with the provided responses ranging through 'every day,' to 'several times in a week,' 'once or twice in a week,' 'once or twice in a month,' and 'never.' This question was followed by four additional items that inquired about the intensity with which the respondents followed political news via four different media outlets: TV, radio, newspaper, and the internet. The overall score of measuring political information was drawn from initially reverse-scoring all five items and subsequently obtaining the mean of all of those items.

#### *3.3.3.4 Political Interest*

Political interest was measured through a four-point scale of interest in politics or governmental issues. I gauged the extent to which respondents were interested in politics or in governmental issues in general by employing two items. The first was: "How interested are you in politics or governmental issues in general?" Following that, I also asked: "How often do you talk/discuss government issues with other people (family, neighbour, colleague, etc.)?" Because of the nature of these questions, I needed to reverse-score the two items. Finally, I added the score of the two items and divided the result by two, leading a four-point scale of political interest, ranging from not interested at all (1) to very interested (4).

### *3.3.3.5 Political Trust*

To measure political trust in a comparable way, I used a four-point scale constructed from seven items that focused on respondents' level of trust in seven key political institutions. The main question was: "We will mention several institutions. Please rate your level of trust in these institutions: do you possess high trust, moderate trust, little trust, or no trust at all? For each one, please tell me how much trust you have in them: President, People's Consultative Assembly (MPR), People's Representative Council (DPR), judiciary, the armed forces, police, and General Electoral Commission (KPU)." Again, all seven items above were reverse-scored, added up, and subsequently divided by seven, establishing a four-point scale of political trust, ranging from no trust at all (1) to high trust (4).

## *3.3.3.6 Political Participation*

For this study, a scale of electoral participation was constructed from a pair of items that indicated whether or not (and if not, why) a citizen participated in the last two legislative elections. The first question was "Did you vote during the 2009 legislative election?" The second was "Did you vote during the legislative election on April 9, 2014?" Each of these two items provided a number of answer choices, varying from 'Yes, I voted,' 'No, I didn't vote,' 'couldn't vote,' to 'refused to vote.' I established turnout as a binary indicator, taking on a value of 1 if the individual voted in the 2009 or 2014 elections, and 0 if the person didn't vote in both elections. The overall political participation score is an additive 0–2 index, based on the two items described above.

# *3.3.3.7 Democratic Support*

This study used a single standardised item, which was adopted from similar international measures on support for democracy. The question was: "Do you strongly disagree, disagree, agree or strongly agree with this following statement? Compared to other form of governments, democracy is the best form of government for our type of country." As in the original variable, the result is measured on a four-point scale, varying from strongly disagree (1) to strongly agree (4).

# 3.4 Results and Discussion: The Determinants of Vote Buying: Who Gets Targeted?

Having introduced key hypotheses and the measures to verify or falsify them, it is now time to analyse the results regarding the determinants of vote buying based on the post-legislative election survey in 2014. In order to assess the impact of covariates at the individual level, I first present the results from the regression analysis, using individuals' selfreported experiences of vote buying. At the end of the section, I highlight the most striking factor emerging from as being a significant determinant in the targeting of vote buying: party identification. The summary statistics for both dependent variables and covariates are listed in Table 3.1.


**Table 3.1** Descriptive dependent variables and covariates

Table 3.2 shows the regression results with respect to vote buying. The dependent variable is the vote-buying index based on responses to all components of this practice (see Chap. 2). I estimated the model using linear regression analysis.6 Model 1 of Table 3.2 allows us to assess the role of the modernisation hypothesis and other socio-demographic variables to

6To be accepted as a multiple linear regression model assumption, it is first necessary to test the classical assumption, which at least includes a normality test. Based on the results of the normality test I conducted, as can be seen in Appendix D, the residuals weren't normally distributed. Accordingly, I needed to conduct a logistic regression because it doesn't need many of the classic assumption tests (including a normality test) that are required in linear relationships between the dependent and independent variables. In the logistic model, the dependent variable is a dichotomous measure that takes a value of 1 if the respondent received offers of vote buying once/twice, several times, rarely, quite, or very often. The results from both the logistic and linear models are strikingly similar (see Appendix E), suggesting that the linear regression in this case fits just as well as the logistic regression. In this chapter, I decided to report the linear regression model rather than the logistic model for two reasons. First, given that the linear and logistic regression analyses end up with indistinguishable results, the linear model is justifiable. Second, compared to the logistic regression, the linear estimates are easier to interpret (Hellevik, 2007).


**Table 3.2** Linear regression analysis of the determinants of vote-buying targeting Note: \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001; b = Unstandardised coefficients; Beta = Standardised coefficients explain the likelihood that an individual is targeted for vote buying. Contrary to the assumptions of numerous scholars, indicators normally advanced to support modernisation arguments about clientelism generally fail to predict the probability that respondents reported being offered benefits in exchange for their votes during the 2014 legislative election. Individual differences in terms of level of income, education, and sociogeographical location didn't make a difference in determining the likelihood that a respondent would be targeted.

Model 1 of Table 3.2 clearly demonstrates that the estimates of these three important measures of modernisation theory (level of income, education, and socio-geographical location) aren't statistically significant in the strictest sense.7 When I introduced controls for civic engagement (Model 2) and for a set of political attitudes (Model 3), the magnitudes of these variables remained insignificant.

As indicated above, the insignificance of income and educational factors as well as socio-geographical location tends to contradict the claims of vote-buying scholars who use modernisation theory in their work. Scholars have long suggested that those with lower levels of education will more likely be engaged in quid pro quo exchanges before and during elections (e.g. Kitschelt, 2000; Brusco et al., 2004; Çarkoglu and Aytac, 2015). Equally, it has been claimed that respondents with more education will be less vulnerable to such practices since "as education increases, the life prospects for individuals also increase through better employment and higher incomes" (Sugiyama and Hunter, 2013: 51). Similarly, it is also proposed in the literature that voters who reside in rural areas are more likely to sell their votes (Hicken, 2007a; Jensen and Justesen, 2014; Finan and Schechter, 2012; Vicente, 2013). However, my survey results show that none of these arguments are supported in the case of Indonesia.

More unexpectedly, Model 1 of Table 3.2 also shows that among the socio-demographic variables included in the equation, three variables have a significant relationship with vote buying: gender (male), age, and religion (Muslim). When adding the control variable of civic engagement, as in Model 2 of Table 3.2, the coefficient of male loses statistical significance. In contrast, the magnitude of the age coefficient increases in specification

<sup>7</sup> It must be noted that bivariate statistics reveal that vote buying has apparently a significant association with the level of education. But this relationship is spurious given the nonsignificance of the education variable when controlled by other variables. For further discussion on the bivariate results, see Appendix C.

when civic engagement is put into the model. Meanwhile, the coefficient of 'Muslim' remains significant at about the same level of strength—and in the same direction—when the element of civic engagement is included. This confirms the bivariate model suggesting that younger citizens8 and Muslims reported statistically significant higher frequencies of vote-buying offers (see Appendix C). However, when I entered a full set of controls (as in the final column of Table 3.2), the only socio-demographic variable that remained statistically significant was that of being Muslim. This significance was at the 0.05 levels. Put differently, being a Muslim significantly predicts a greater probability that a respondent reported electoral handouts, even when a full set of controls was included.

The consistency of this relationship requires investigation. There are two possible explanations. First, there is ample evidence that Muslim candidates and recipients justified vote buying using religious reasoning. Many candidates, either from Islamic or non-Islamic parties, packaged their gifts in religious terms, for example, by describing them as alms (*sedekah*). Accordingly, the Muslim recipients might not feel that the cash handouts they received were a form of morally questionable 'money politics' but instead viewed them in terms of a 'moral economy of gift giving' (Aspinall et al., 2017: 4). Second, vote buying appears to be endemic among Muslims because traditional patron-client networks—through which candidates' benefits can be delivered—are more prominent in Muslim communities. In particular, many candidates used various deeply entrenched informal social institutions, such as Islamic boarding schools (*pesantren*), mosques, or religious gatherings (*majelis taklim*). Importantly, such loose informal networks don't always constitute themselves through official membership and are thus outside of the civic engagement linkages that respondents reported when asked about their membership in social organisations.

While loose social interactions in Muslim society may have played a role in increasing their exposure to clientelistic practices, the overall effect of

8One potential explanation for this is that young adults are considered to be politically unsettled and thus have weaker political preferences and voting intentions. Opinion polls show, for example, that their intentions to show up at the polls aren't as strong as those of the older voters. According to Indikator's exit poll in the 2014 legislative elections, the proportion of registered 26- to 40-year-olds who turned out to cast their ballot was lower than that of other age groups. These hesitant young voters may create a strong incentive for voteseeking politicians to induce them with benefits. See Indikator's exit poll material release at http://indikator.co.id/uploads/20140411204045.Hasil\_EP\_Pileg\_2014\_Update.pdf.

civic engagement on vote buying isn't significant. The regression analysis, as shown in Model 2 of Table 3.2, confirms the bivariate statistics that the substantive effect of civic engagement—measured by membership in social organisations—is statistically insignificant in the strictest sense.9 As noted above, much of the literature on electoral clientelism suggests that networks of civic engagement explain the likelihood of being targeted with electoral incentives (Faughnan and Zechmeister, 2011; Callahan, 2005; Brusco et al., 2004; Cruz, 2014). But my analysis suggests that civic engagement in Indonesia has little impact on the individual's propensity to be so targeted. Individuals' involvement in social organisations makes no difference in the probability of them being exposed to vote buying.

What about political attitudes? Three observations stand out in Model 3 of Table 3.2. First, across all components of political attitudes, party identification has a statistically significant and positive relationship with vote buying. The multivariate analysis shows that this partisanship isn't spurious and remains significant and relatively stable regardless of civic engagement, political attitudes, and demographic and socio-economic factors. The effect of party closeness is very large (*p* < 0.001). Extant literature suggests that party affiliation encourages citizens to become politically active. Dalton (2016: 8) found that in the US, people with higher party identification are more likely to persuade others, distribute propaganda hats, shirts, and posters, participate in campaign events, or donate to a party's candidate during campaigns. Electoral turnout, Dalton continues, was 26% higher among strong partisans than among independents. Verba and Nie (1972: 219–220) claim that partisan identification mobilises political activity among lower-status citizens who might otherwise be inactive. The same holds true in other established democracies. Voters with strong party identification voted at a higher rate in the 2009 German Bundestag elections. In addition, they were several times more likely to get involved in campaign rallies and were about twice as likely to try to influence other voters (Dalton, 2016: 8). What stands out about party identification in Indonesia, by contrast, is its strong correlation with experience of vote buying. In crude terms, partisan voters in Indonesia tended to be 'money grubbing' and are significantly more likely to be recipients of material benefits.

Second, defying expectations, the model also shows that political interest is found to be statistically very significant (*p* < 0.001) in explaining the

<sup>9</sup>For further discussion on the bivariate results, see Appendix C.

propensity of being offered vote buying. The original relationship between political interest and vote buying remains substantial and significant, even when controlled by all variables in the model. Thus, instead of reducing the individual's propensity to engage in clientelistic practices, political interest serves—in the Indonesian case—as a breeding ground for vote buying. The dominant view among political scientists is that a good citizen in a democratic polity is a citizen interested in and well informed about politics, with strong and stable preferences (e.g. Van Deth, 1983, 1989). The regression analyses of the Indonesian data, however, shows that political interest is positively correlated with a distinctly nondemocratic practice, that is, vote buying.

There are two possible explanations for this rather counter-intuitive finding. First, it is plausible that in the Indonesian case, political interest is an indicator of political alienation. Arguably, those who are interested in politics actually desire to participate in a democratic manner, but that desire is undermined by their perception that the political process doesn't operate justly. This perception, then, makes them increasingly pragmatic at election time (this interpretation is supported by the analysis of the efficacy measure below). Another possible explanation for the link between political interest and vote buying has been suggested by Guardado and Wantchekon (2014: 7). They propose that exposure to electoral handouts is endogenous to voters' political interest. Based on their theory, if voters are highly interested in politics, they are likely to get involved in more political activities than others. This, in turn, increases their likelihood of being offered a reward by political operatives they come into contact with.

Finally, consistent with prior expectations (and as indicated above), regression results reveal that political efficacy stands out as a strong predictor of vote buying. The variable of efficacy presents a negative coefficient and is in the expected direction, implying that less efficacious individuals are more likely to receive offers of vote buying. The reverse is also true: those with higher efficacy are less likely to engage in such practice. Of course, this finding is neither new nor surprising, given that the existing literature on clientelism has long suggested such a relationship (e.g. Banegas, 1998; Kerkvliet, 1991; Schaffer and Schedler, 2007). By contrast, other political attitudes have little effect in determining vote buying in Indonesia, including support for democracy or political (i.e. electoral) participation, which I had earlier hypothesised would show correlations with vote-buying experience. Indonesian respondents supporting democracy aren't more or less likely to engage in vote buying than those who don't, and electoral participation is no reliable predictor of such engagement either.

Having established the regression results with respect to vote buying at the national legislative election, we now need to highlight the most salient issue found as being significant in such electoral setting. The finding is clear: party identification has a strong, consistent, direct, positive, and highly significant effect on an individual's propensity to receive offers of vote buying. This effect remains significant, regardless of socio-economic factors, civic engagement, and political attitudes. In other words, a voter's partisanship consistently and significantly predicts the likelihood of being offered vote buying in Indonesia. Other factors stand out as well—such as being Muslim—but in the largest Muslim nation on the globe, this isn't a theoretically challenging finding.

On the other hand, the finding, that party loyalists are the primary targets of vote buying, obviously contrasts with the swing-voter model, which proposes that parties or candidates will not waste their vote-buying efforts on already locked-in partisans (see Chap. 1). As I will argue in the following chapter, the underlying logic behind candidates' decision to target loyalists is that they are safe bets, or a good return on a limited investment. By contrast, from the candidates' perspective, targeting non-partisans is a risky enterprise. Hence, investing in party loyalists is a function of candidates' attempts to manage and mitigate electoral uncertainty. A widely held view among candidates and brokers is that because electoral competition has become increasingly competitive after the introduction of the open-list PR system, and in an environment where electoral participation is optional and the ballot is secret, it is important to target 'loyal' or 'base' voters, and of course party loyalist form a large subset within this group.

#### 3.5 Conclusion

At the beginning of this chapter, I developed a number of hypotheses on who is most likely to be targeted by vote-buying efforts. These hypotheses were drawn from the main streams of the vote-buying literature. Let us now review systematically whether these hypotheses hold. It is clear that factors derived from modernisation theory have little effect in determining an individuals' propensity to experience vote buying in the national legislative election. One of the most common measures of modernisation theory is education. Some assume that better educated voters are less likely to engage in vote buying, partly because they understand that the practice undermines democracy and partly because they are economically better off. But in the case of Indonesia, there are no significant differences by educational level in the receipt of electoral incentives.

Another surprising result is the insignificance of poverty. One of the strongest findings in prior research on clientelism is that poorer voters are more susceptible to vote buying because even small transfers are valuable to them. In Indonesia, however, vote buying has no significant relationship with poverty. Similarly, vote buying in the country didn't differ between rural and urban regions. Additionally, this study explored whether civic engagement can explain the likelihood of being offered vote buying in the run-up to the 2014 legislative election. This hypothesis, however, is mostly denied in the case of Indonesia. Involvement in social organisations didn't make a difference in terms of whether citizens received offers of vote buying or not.

With respect to specific hypotheses relating to political attitudes, I expected, drawing from the swing-voter model, that party identification levels would be negatively associated with the likelihood of experiencing vote buying. In other words, I hypothesised that the more non-partisan voters were, the more likely they would be to receive vote-buying offers. In the case of Indonesia, however, the opposite turns out to be true: partisan voters are proportionally more likely to receive vote-buying offers than non-partisans. In terms of political efficacy, I expected to find that feelings of political alienation increase the likelihood of exposure to clientelist exchanges (Reef and Knoke, 1999: 414). Using Indonesia's 2014 legislative election, I verified this hypothesis. Another element of political attitude dimension is political interest. I expected to find that the more uninterested a citizen in politics, the more vulnerable they would be to vote buying. Against expectations, instead of interest in politics predicting less exposure to vote buying, it increased it. Interest in politics significantly predicts the probability that someone would report vote buying. Another important political attitude is political trust. People with lower political trust are believed to be more unlikely to trust democratic institutions, and they are therefore more likely to build transactional relationships with political parties or candidates (as indicated above, this is similar to arguments surrounding efficacy, but political trust is a more abstract concept). If this claim is true, I expected to find that citizens with less trust in political institutions to be more likely to engage in vote buying. In the Indonesian case, this hypothesis isn't falsified since the bivariate and multivariate models found no evidence; there was simply no relationship to clientelism. Further, there is also no statistically significant relationship between vote buying and electoral participation, political information, and support for democracy.

Given the strong evidence for the importance of party identification in predicting the likelihood of a voter being targeted by vote buying, the remainder of this book focuses on how exactly such party-based partisanship affects electoral patronage distribution. The finding that party affiliation significantly attracts benefits raises further puzzles. First, why do parties and candidates disproportionately target party loyalists with their favours? As indicated, this finding runs counter to prior research done by advocates of the swing-voter argument, who suggest that such persons are already captive voters and will support their party no matter what. Second, if partisans are indeed the preferred vote-buying target, how feasible is it for candidates to win only by targeting such voters? How many partisan voters are there, and how does their number affect further efforts to theorise vote-buying patterns in Indonesia? The next chapter addresses these questions.

#### Bibliography

Books, Journal Articles, and Unpublished Papers


———. (2007a) How Do Rules and Institutions Encourage Vote Buying? In: F.C. Schaffer (ed) Elections for Sale: The Causes and Consequences of Vote Buying. Colorado: Lynne Rienner Publisher, Inc.

———. (2007b) How Effective are Institutional Reforms? In: F.C. Schaffer (ed) Elections for Sale: The Causes and Consequences of Vote Buying. Colorado: Lynne Rienner Publisher, Inc.


———. (2007) The Demise of Clientelism in Affluent Capitalist Democracies. In: H. Kitschelt and S. Wilkinson (eds) Patrons, Clients, and Policies: Patterns of Democratic Accountability and Political Competition. Cambridge, UK: Cambridge University Press.


Van Deth, J.W. (1983) The Persistence of Materialist and Postmaterialist Value Orientation. European Journal of Political Research 11: 63–79.

———. (1989) Interest in Politics. In: K. Jenings and J.W. Van Deth (eds) Continuities in Political Action. New York: Walter de Gruyter.


**Open Access** This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

# Do Candidates Target Loyalists or Swing Voters? Beyond the Core- Versus Swing-Voter Debate

In the previous chapter, I have identified the typical characteristics of voters who were most likely to experience vote buying. Among a number of findings, the most consistent and intriguing was that a high level of party identification, that is, self-reported closeness to one of Indonesia's political parties, is a significant predictor. In this chapter, I turn in more detail to the interplay between party identification and vote buying. This linkage clearly relates to the core- versus swing-voter debate that has been the primary concern of much of the comparative literature on vote buying. At the centre of this debate is one key question: given the budgetary constraints candidates face, how and, especially, to whom do they distribute gifts to optimise their electoral prospects? As I will show in the following pages, most candidates and brokers repeatedly claim that they target partisan, loyalist voters. And as indicated in the previous chapter, voters with high levels of party identification are indeed—in *relative* terms—more likely to be targets of vote buying. But as the discussion below reveals, the number of party loyalists in Indonesia is small, leading to a situation in which the vast majority of vote buying—in *absolute* terms—happens among non-partisan voters. Hence, the main task of this chapter is to unpack this complex interrelationship between party identification, broker and candidate strategies, and the observed practice of vote buying on the ground.

Given the centrality of party identification in attracting benefits, this chapter begins with descriptive data about the level of mass partisanship and its distribution across political parties in Indonesia. Then I consider whether the substantial effect of partisanship on vote buying reflects a case of reverse causality: did such voters become the target of vote buying because they had professed greater ideological proximity to the party? Or, did electoral handouts come first and then predict likelihood of being close to the party? The next question I discuss is variation across parties, and whether support for any particular political party is most closely identified with vote buying. Next, I examine the core- versus swing-voter models, highlighting evidence from a survey of politicians and brokers as well as qualitative data that seem to point to a tendency towards corevoter strategies among election practitioners. The chapter then discusses the apparent paradox that although in *relative* terms party loyalists are more likely to be targeted, and politicians and brokers confirm that such voters are their primary targets, in *absolute* terms most vote buying occurs among non-partisans. This finding leads to further puzzles with regard to whether failed targeting took place or whether there are other factors that can explain this pattern. The chapter, therefore, ends by setting up the key questions for the following chapters to tackle.

### 4.1 Political Party Partisanship in Indonesia

At the outset, it is essential to show the aggregate level of self-reported party identification in Indonesia based on my post-election survey of voters in April 2014. While there is much divergence of opinion on the nature and measurement of party closeness (Blais et al., 2001; Greene, 2002), this study measures the degree of partisanship regarding a political party by using the three items introduced in Chap. 3. Through the first measure, respondents were asked whether they feel close to any political party. In my post-2014 legislative election survey, only 14.9% (herein we round up to 15%) nationally reported having such closeness—a low figure by international standards. As discussed in Chap. 2, there were around 187 million registered domestic voters in the 2014 legislative election. Hence, the 15% would mean an estimated 28 million voters nationwide felt close to a party.

For the purpose of this study, I categorised as 'non-partisan' those respondents who either gave a straight negative response or couldn't answer this question. Consequently, the number of non-partisans is—at 85%—extremely

**Fig. 4.1** Distribution of partisanship across political parties (%). Source: My post-election survey, 22–26 April 2014

high, constituting the vast majority of the electorate, or about 159 million voters. As will be further demonstrated in the following section, this pattern was confirmed in a series of nationwide surveys during the run-up to 2014 elections.

Those who answered the opening question with 'yes' were requested to name the specific party they feel close to. Figure 4.1 shows that mass partisanship in Indonesia varies widely across party distribution. Of those expressing partisanship, a quarter felt some degree of attachment to PDI-P. Following PDI-P was Golkar with 21.8% and then Gerindra with 13.5%. Thus, among those identifying partisanship, more than 60% of respondents felt close to one of the three largest parties. At the other extreme of the spectrum, we find parties with almost no partisans: PBB (0%) and PKPI (0.6%). To some extent, the distribution of party loyalty reflects the distribution of votes in the 2014 parliamentary elections. The big three of partisan identifiers, PDI-P, Golkar, and Gerindra, were placed in the top three spots and in the same order in the election results.1 However, it is

<sup>1</sup> In 2009, the order of these parties weren't the same. In the February 2009 survey of LSI, among those reported being close to a party (19.7%), 18.2% felt close to the Democratic Party, followed by PDI-P (18.1%), and Golkar (17.3%). This further bolsters my argument as I will subtantiate later that even so-called 'partisans' may not be particularly loyal to their party.

important to note that the vote totals for all parties were several times larger than the number of voters who expressed allegiance to those parties.

In the third and final measure, those who reported being close to a party were asked to rate the strength of this affiliation on a three-point scale: "How close do you feel toward the party?" Those who said 'very close' to the party were classified as *strong partisans*. Those who replied 'quite or fairly close' to the relevant party were categorised as *moderate partisans*, while those who reported 'a little close' to a party were classified as *weak partisans*. Among the 15% of the respondents who felt close to any party, moderate partisans were the largest subgroup (58%), with weak partisans (23.1%) and strong partisans (17.4%) constituting much smaller segments. Note that those who reported having varying levels of closeness to a party are distributed across political parties.

The low scores of mass partisanship as discussed above leave a number of unanswered questions: if there are so few voters with clear and declared loyalty towards a particular party, why were candidates very keen to target such voters? In the same vein, considering the large proportion of nonpartisans with a greater potential to change their voting decisions if given benefits (Lindbeck and Weibull, 1987; Dixit and Londregan, 1996; Stokes, 2005), as thoroughly discussed in Chap. 1, why did political machines profess that these voters were secondary to their vote-buying strategies? And finally, given that in total numbers more non-partisans than partisans experienced vote-buying attempts, does that mean that candidates and brokers misdirected their vote-buying operations? But first, I will discuss whether or not voters' partisan closeness is entirely a result of their receiving electoral rewards.

# 4.2 Is Party Identification Endogenous to Benefits?

The finding that party affiliation is a significant predictor of vote buying raises an important question about the potential of reverse causality. It may be argued, as Diaz-Cayeros and his collaborators (2012: 159) have done, that the linkage between partisanship and vote buying is a case of 'conditional partisan loyalty,' that is, it is strongly influenced by distribution itself. They argue that the driving factor behind politicians' tendency to funnel benefits to their own supporters is the endogeneity of partisan loyalties to material inducements (Diaz-Cayeros et al., 2012). Stokes and her colleagues (2013: 54) have indicated the same potential endogenousloyalty problem, aptly summarising the problem as follows: rather than voters' political preferences attracting handouts, these electoral incentives may cause people to identify with and support the party that gives them.

We must discuss two important caveats when dealing with the probability of an endogeneity problem (Stokes et al., 2013: 54–55). Since this study relies on survey data, the problem of reverse causality might be the result of a measurement error. To minimise this potential bias, I developed survey instruments for measuring political party partisanship based on those proposed by the Comparative Study of Electoral Systems (CSES). These measurements are plausibly proven to attract answers that aren't caused by a respondent's receiving particularistic rewards. Additionally, I put the question about respondents' feelings about parties at the beginning of the relevant questionnaire section to avoid the possibility of being contaminated by later questions on whether they have received a gift or social benefits from a party or candidate. By doing so, this question order is expected to be statistically independent of respondents' potential responses that could take the rewards they receive into account. Nonetheless, this methodological survey approach doesn't yet fully address the probability that respondents could conceptualise party loyalty as conditional on the rewards they received, rather than fixed (Diaz-Cayeros et al., 2012: 23).

To test for endogeneity bias, this study presents very simple statistics. Figure 4.2 suggests that party loyalty in Indonesia isn't conditional on receiving offers of vote buying. If party loyalty was endogenous to distribution of rewards, we might expect that partisan party alignment would increase parallel to the gradual intensification of vote buying as an election approaches. The tracking polls I conducted in the lead up to the 2014 elections allow us to examine whether this occurred. The square dotted line traces the percentage of respondents who feel close to political parties, and the long dashed line pinpoints the percentage of respondents who experienced offers of electoral rewards. In order to detect potential reciprocal effects of vote-buying transactions on a recipient's 'closeness' to the party, we need to track the two lines over time. The square dotted line shows that the percentage of people who reported being targeted with benefits started at a low point in January 2014 but increased over the course of the campaign, from 4.3% in January to 8.1% in late February to early March, 10.7% in late March, and 25% in the few days leading up to the election in April.

If the endogenous loyalty thesis were correct, party identification levels should have risen in concert with the dramatic increase of vote-buying incidents in the weeks leading to the poll. However, the square dotted line

**Fig. 4.2** Party identification and vote buying approaching the 2014 election (%). Source: The January, February–March, and late March 2014 data were taken from my pre-election surveys, while the April 2014 numbers were drawn from my postelection survey (see Appendix A)

clearly indicates that partisan loyalties around the 2014 elections were relatively stable, ranging from 12.9% to 14.9%. Put simply, party identification was virtually unaffected by short-term electoral incentives, providing evidence that the effect of voters' partisanship on their individual probability of being targeted by vote buying isn't endogenous, and that a voter's party loyalties in Indonesia aren't a function of his or her clientelist interactions with the party or its candidates. This finding is in line with the claims of early proponents of the concept of party identification, who defined party identification as a sense of personal, psychological attachment (Campbell et al., 1960) and viewed party identification as more of an identity than an opinion (Larcinese et al., 2012: 3). Partisanship, therefore, is independent of short-term factors such as vote buying.

# 4.3 Variations in Party Identification by Party and Vote Buying

Having ruled out the possibility of reverse causation, it is now important to examine whether partisanship towards any particular political parties is most closely identified with vote buying. Using data from my post-election survey in April 2014, it allows for a broader comparison across legislative, local, and presidential elections. I first recoded party identification into new variables consisting of partisan identities in each party. In every partyspecific partisanship, I transformed an existing variable into four categories reflecting varying degrees of partisanship. For instance, the partisanship to NasDem variable ('Nasdem partisan') was recoded into four categories: 0 refers to respondents who didn't identify with NasDem, including those who were ideologically indifferent or opposed to the party; 1 represents weak partisans; 2 points to moderate partisans; and 3 indicates strong partisans of NasDem. The same approach was used for other partisan variables in each political party. Afterwards, to measure the strength of a linear correlation between the two variables, each partisan identity was correlated with vote buying.

Table 4.1 shows important variations among core constituents of all parties. If we refer to the vote buying in legislative elections, partisans of the three big parties, PDI-P, Gerindra, and Democrats (as well as PKS), are more likely than others to be targeted for vote buying. Similarly, Golkar partisan affiliation has a markedly strong relationship with the overall vote-buying


**Table 4.1** Correlations (Pearson's *r*) between party identification and vote buying

(a) Cannot be computed because at least one of the variables is constant (no cases)

\*\*\**p* < 0.001, \*\**p* < 0.01, \**p* < 0.05

index. As discussed in Chap. 2, the vote-buying index is a composite statistic of vote buying in legislative elections and vote buying in presidential and local executive elections. In general, these statistical results correspond with findings from a related but differently worded question already analysed in Chap. 2: when asked which party or candidates offered goods or money, the most frequent answer was Golkar (32.2%). Following Golkar are PDI-P (26.5%), Gerindra (25%), and the Democratic Party (18.4%).

Alternatively, in order to examine partisanship with which party increases the likelihood of being offered benefits, I correlate the variables of party identification with clientelist exchanges in each political constituency. I conducted such correlation because there are predictable variations in the level of party identification across political constituencies and people who voted for certain political party weren't necessarily close to that party. The results, however, are generally similar to those produced by the first technique.

Table 4.2 reveals a clear positive correlation between party identification and clientelism in some political constituencies. Party-based identification among those who voted for PDI-P, Golkar, and Democratic Party in the 2014 legislative election correlates with greater


**Table 4.2** Correlations (Pearson's *r*) between party identification and vote buying in each group of political constituencies

(a) Cannot be computed because at least one of the variables is constant (no cases)

\*\*\**p* < 0.001, \*\**p* < 0.01, \**p* < 0.05

likelihood of engagement in vote buying, defined in this study based on a combination of responses to a battery of questions that inquired about respondents' exposure to such practices in the legislative elections as well as national and local executive elections (see Chap. 2). As can be seen in the third column of Table 4.2, the magnitudes of the relationship are strong since their substantive effects reach the 0.05 significance, or even more. Similar (but not entirely congruent) with the first model, when we restrict our analysis only to vote buying in legislative elections, partisan attachments among those who voted for PDI-P, Golkar, PKS, Hanura, NasDem, and PPP (thus more than half of all parties in the 2014–2019 parliament) are found to be significantly correlated with vote buying.

Overall, despite persons expressing party loyalty attracting benefits, we must be especially attentive to variations. Such relatively consistent findings between the two statistical techniques reinforce the notion that partisanship towards some parties doesn't automatically make such partisans a vote-buying target. Only where a positive, direct, and significant correlation with such handouts exists can we tie particular party-based partisanship to a higher likelihood of being targeted with clientelist techniques.

### 4.4 Testing the Models of Distributive Politics

#### *4.4.1 Evidence from Surveys of Local Politicians and Brokers*

In the preceding sections, I discussed survey data on the interplay between party identification and clientelism and addressed a number of methodological problems in analysing this interplay. This has put me in a position to now further review the debate about core and swing voters against the background of findings on the influence of voters' partisan predispositions.

So far, our examination about the conflicting strategies of core- or swing-voter targeting has only dealt with the demand side of vote buying, namely the voters. However, the supply side is equally important: how do candidates and brokers view the core- versus swing-voters dilemma? In order to take additional steps to explore whether candidates and brokers target core or swing voters in Indonesia, this section primarily draws from the survey of low-level politicians and brokers in four provinces (West Sumatra, Central Java, East Java, and North Sulawesi) I conducted in September and October 2014 and which was described in Chap. 1 and is explained in greater detail in Appendix B. This survey asked respondents numerous direct questions about their clientelistic practices. The total sample was 1199 respondents consisting of 299 provincial and district legislators and 900 brokers (known locally as 'success teams') who worked for them in the 2014 elections.

Stokes and her colleagues (2013: 31) argue that, faced with limited budgets, political parties and their candidates don't waste resources on targeting core voters with material inducements, assuming that they are likely to vote for them anyway. At the same time, however, rewarding voters who are ideologically distant from or opposed to the party or candidate is also considered wasteful (Stokes, 2005). Accordingly, candidates are assumed to reach out to swing voters in the middle of these two extremes, that is, uncommitted voters who will reciprocate with votes for any kind of gift. Reflecting these dynamics, political machines usually divide locales into three categories: party base, party opponent base, and locations of swing voters. It is important to note that the distinction between 'party base' and 'individual candidate base' is often blurred in the Indonesian context; I return to this point in Chap. 5. It is in this context that respondents in my survey addressed the question of whether they distributed largesse in party base, swing voter, or opponent base areas.

The possible overlap of party with personal networks notwithstanding, the results of the survey show that politicians and electoral brokers claimed to have distributed more benefits in their party base areas than in other territories (Fig. 4.3). This finding is, of course, inconsistent with the expectation inherent in the swing-voter hypothesis, and instead seems to confirm the notion of core-voter targeting proposed by Nichter (2008) and others (Cox and McCubbins, 1986; Diaz-Cayeros et al., 2012; Stokes et al., 2013). I didn't find considerable heterogeneity by province. Compared to their operatives or brokers, candidates exhibited even greater inclinations to funnel resources to locations where their parties performed well in the past. It is worth noting, however, that in the context of an open-list system—as I will further discuss in Chap. 5—contenders win seats not only by defeating candidates from other parties. More importantly, they have to beat co-partisans from their own party—and this makes the party base a particularly contested field. Consequently, party nominees 'individualise' the party base by distributing more resources to party supporters in an effort to maximise their individual votes from them. The implication is clear: they need to outdo their internal rivals in terms of patronage distribution.

**Fig. 4.3** Party base is the preferred target (%) (All sample here represents a combination of local MPs and brokers. However, as indicated earlier, the proportion of brokers surveyed is much larger than the proportion of local politicians interviewed. See Appendix B.). The survey question used is: "During the last 2014 legislative election, how did you distribute largesse in order to get votes?" Source: My survey of low-level politicians and brokers, 30 September–25 October 2014

The same trend emerged from a measure that explored the priorities of resource allocation to different broker teams. The wording of the question was as follows: "There are two different success teams working in two different areas. One area is known as party base area, while the other is known as the swing competitive area for political parties. Between these two success teams, which one will get more money and logistical support from the party or candidate?" Figure 4.4 reveals that the majority of local MPs and brokers preferred to allocate more resources to campaign teams working in party strongholds than in swing areas. Again, the figure is strongly suggestive that Indonesian candidates prioritise mobilising their own party supporters: what Nichter (2010) famously called 'turnout buying,' and Schaffer and Schedler (2007: 25) briefly refer to as 'participation buying.'

In order to further consolidate the finding, I inserted an additional question on this matter in my broker survey: "If you distributed envelopes/staple goods package, which type of voters did you prioritise?" Three choices were provided: (1) Voters who regularly vote for the party I support; (2) Voters who regularly vote for another party; and (3) Voters whose voting behaviour cannot be determined. Over 75% of brokers

**Fig. 4.4** Party base receives more resources (%). Source: My survey of low-level politicians and brokers, 30 September–25 October 2014

admitted to targeting voters they thought of as being regular party supporters. Only one-fifth favoured swing voters, while only 2% preferred the supporters of rivals. Asked in an open-ended question to specify the main reason for prioritising such partisan, loyalist voters, the most typical response was that these voters were relatively certain targets. However, I found a significant variation to this trend in West Sumatra (as previously noted, an area of relatively low vote buying), where non-partisan voters were disproportionately targeted for vote buying. This stood in contrast to Central Java, East Java, and North Sulawesi, where brokers favoured those perceived as party loyalists over ideologically indifferent voters.

In short, the evidence emerging from the various surveys suggests that in Indonesia, brokers and candidates tend to direct patronage distribution flows during elections towards voters with strong party-based loyalties. As I will discuss in the following section, the logic behind this inclination to prioritise loyalists over swing voters is largely to reduce risk. Core supporters are seen as less risky, more responsive, and predictable targets, which is, at first glance, consistent with the core-voter argument.

#### *4.4.2 Reasons Behind the Tendency to Favour Core Voters*

#### *4.4.2.1 Double-Layered Risk Aversion*

Having stated the favoured tendency among political actors, I now turn to delve more deeply into explanations for why candidates and brokers preferred to channel benefits to party loyalists rather than to uncommitted voters. In order to explain the logic behind their proclaimed core-voter strategies, I draw from the survey of local politicians and brokers and indepth interviews with national politicians. After posing the hypothetical questions analysed above, I asked brokers and candidates for a spontaneous response to an open-ended question explaining their choice. As in Fig. 4.5, the respondents who reported expressing a strong desire to target loyalists (recall that this was the majority of politicians and brokers) did so due largely to reduce risk as core supporters are seen as less risky, more responsive, and predictable.

This finding was echoed by many candidates I interviewed during my 13 months of fieldwork in Indonesia. Many stated that they were mostly concerned about the threat of opportunistic defection by voters when engaging in vote buying. In order to minimise uncertainty over whether or not the recipient would repay the supplied benefits with votes, riskaverse candidates tend to target their assumed core supporters whose support can—according to the expectations of campaign organisers—be maximised by increasing turnout. In short, the strong preference to favour

**Fig. 4.5** Why do success team members target loyalists (left panel) or uncommitted voters (right panel)? (%). Source: My survey of low-level politicians and brokers, 30 September–25 October 2014

core supporters reflects the search for certainty among both candidates and intermediaries. Thus, this study confirms propositions by early corevoter theorists (Cox and McCubbins, 1986: 379) who argued that political actors are inherently risk-averse and are reluctant to shower benefits on swing voters because many such voters might defect.

This analysis is supported by my targeted interviews with national-level politicians. One prominent national politician from the traditionalist Muslim party PKB (National Awakening Party) which was founded by Nahdlatul Ulama (NU), Indonesia's largest traditionalist Muslim organisation, admitted that most of his campaign benefits were distributed to his party's bases. He explains:

I asked my teams to deploy more resources to traditionalist *santri* (devout Muslim) bases because they are the most reliable voters with the highest electoral potential. Those affiliated to NU, the traditionalist *santri* are my party's real constituency. … For me, targeting NU followers that are closely connected historically as well as ideologically to PKB was more appealing and much more certain. (Interview, 18 April 2014)

A successful candidate from the Islamist party PPP (United Development Party) employed a portfolio of strategies that maximised his electoral support by investing in the loyalty of his own party and personal supporters. Asked why he predominantly distributed campaign benefits to loyal supporters, the candidate, who was running in South Kalimantan, where Banjarese were the largest ethnic group, cited a local philosophy: "*Jangankan haruan*<sup>2</sup> *ganal, haruan halus haja bisa meluncat*" (never mind a big fish, even a small fish can escape you). In other words, his view was that loyal supporters are like small fish, whose support is hard enough to secure, while targeting 'big fish' (in this case, swing voters) is much more difficult, needing a lot of effort and resources (Interview, 21 July 2014). Similarly, a candidate from the modernist Islam party PAN explored how potential electoral return rises with an increase in risk. Though uncommitted voters make up a much higher proportion of the electorate, he argued, they are typically associated with high levels of uncertainty. "Pursuing swing voters isn't only a difficult task, but it would also mean overlooking loyal supporters as a captive (*pangsa pasar*) vote," he explained (Interview, 22 April 2014).

2*Haruan* is a species of snakehead fish, an indigenous freshwater fish of Banjar and a common food item among the local populace.

In many electoral territories, there are local equivalents to the Banjarese phrase cited above, and candidates often mentioned them in interviews with me. These phrases generally suggest that individuals operating under conditions of uncertainty should avoid risk,3 with many running along the lines of 'take care of your own' and 'secure your possessions!' (Holder, 1975, cited in Cox and McCubbins, 1986: 383). For instance, a candidate from Hanura focused on turning out his (potential) masses in his electoral district in Central Java so his campaign largesse wasn't wasted (Interview, 14 May 2014). A prominent national leader of PKB, on the other hand, admitted that targeting unknown people was like making a bargain with an uncertain payoff—and costly as well (Interview, 20 April 2014). In emphasising the unreliability of such voters, many candidates characterised undecided voters as '*tidak jelas*' (uncertain) or '*tidak bisa dipegang*' (unreliable, lit. 'cannot be held'). In contrast, they perceived loyal supporters to have what Diaz-Cayeros and his colleagues (2016: 71) called "a high level of adherence" that makes them more responsive to quid pro quo exchanges. In summary, the important driver behind candidates' preference for capturing loyalists was an operational rationalisation of their personal tendency towards risk aversion.

#### *4.4.2.2 The Moral Economy of Vote Buying*

In addition to being rationalised as the most effective and low-risk strategy, some political operators justify and explain the targeting of party loyalists as an act of gift giving—a moral duty, even. As shown in the left panel of Fig. 4.5, when asked why they targeted loyal voters, many of the surveyed candidates and brokers defended it as a reward, a sign of attention or reciprocal action for their supporters. In such cases, as argued by Walker (2014), vote buying isn't defined in terms of economic market transactions; rather, it is a function of complex social relations entangled in a traditional moral economy of exchanging votes for gifts (Aspinall et al., 2017: 4).

In the interpretation of cash handouts or small favours to voters as a ritual gift exchange rather than as an act of blatant vote buying, the recipients don't object to the payment because it isn't seen as a bribe for their

<sup>3</sup>To mention a few: There is a Javanese proverb "*ojo mburu uceng neng kelangan deleg*" (don't chase small fish while losing more valuable goods), or the Sundanese phrase: "*moro julang ngaleupaskeun peusing*" (being tempted by other goods may cause us to overlook our belongings).

votes (Walker, 2014). Instead, they might consider it as part of a moral economy closely associated with elections, or perceive it as signalling "the positive personal qualities of the giver, such as generosity, politeness, responsiveness, and respect…[that] lead citizens to believe that the candidate is good or worthy" (Schaffer and Schedler, 2007: 26). Accordingly, embedding vote buying within social norms of reciprocity helps generate a feeling of obligation on the part of beneficiaries to reciprocate with support (Schaffer, 2007: 193; Aspinall et al., 2017).

Around Asia, there are many terms to describe this phenomenon. The Indonesian trait of *hutang budi* (norm of reciprocity) is somewhat akin to *utang na loob* in the Philippines (Alejo et al., 1996: 84) or *guanxi* in China (Wang, 2013: 4–5). In Indonesia, it is said that once a person has granted us a favour, we should do everything to pay that favour back to him/her, sometimes even at the expense of ourselves. Vote buying fits into this culture of mutual favours, in which candidates are happy to provide small gifts and voters are happy to give their vote. Many candidates typically describe gifts as *uang saku* ('pocket money'). The standard justification is that the 'gift' is given to compensate loyal supporters for the time they lose by going to polling stations (Aspinall et al., 2017: 11). Edi Inrizal, an anthropologist from University of Andalas, Padang, West Sumatra, argued that Indonesians tend to favour indirect references and transactions over blunt treatments of realities. Thus, monetary exchanges need to be packaged in language of morality and generosity. In this context, he mentioned a saying, '*kanai pacak*' (splash out a little money), which is often applied to gift giving in elections. For ordinary people, political power is often equated with big resources and money, so those running for it should distribute small amounts of money to their voters as a cultural token of gratitude (Edi Inrizal, Interview 21 September 2014).

Such practices and the beliefs that legitimate them aren't only limited to voting day, however. When visiting influential religious or other local leaders during the campaign, Hanif Dhakiri from PKB, for instance, didn't come empty-handed, providing them with religiously symbolic goods such as *peci* (associated with Muslim men's hats), sarongs, or even money (Interview, 26 August 2014). Many other candidates did the same thing. They often call these gifts as '*buah tangan*' (keepsake) or '*tanda mata*' (souvenir). Despite the cost that these items incur, there is a widely held view among politicians that allocating resources to loyal supporters isn't as costly as giving to undecided voters. In an attempt to show that core voters are generally the cheapest target, a PAN candidate, running in one electoral district in East Java, explained that

Unlike swing or transactional voters, investing in loyal voters doesn't cost a lot. For Javanese, providing money gifts to loyalists is regarded as *bisyarah* or a sign of love…. It isn't *peningset* … something that binds. At Javanese weddings, the groom gives expensive presents to bind the bride (*peningset*). Instead of binding voters with expensive gifts or big money, *bisyarah* is simple and a cultural tradition. (Interview, 22 April 2014)

Another successful candidate from the Democratic Party, who was running in an electoral district in West Java, confirmed the notion that

*Bisyarah* parallels with cultural values. It isn't only a Javanese, but an Indonesian tradition. If you are seeking people's votes, you have to have good understandings. People who are going to vote for me would be unable to work on voting day. I compensate them for that. It isn't expensive, just keepsakes to 'tie' them so they don't run to cashed-up rivals. If you give your loyal supporters just IDR 10,000 each, that would be enough. Even if a competitor swamped them with big money, let's say IDR 100,000, they would still deliver their votes to us. (Interview, 25 April 2014)

It is noteworthy that candidates and brokers often feel a high level of anxiety if they don't comply with the social norm of gift giving. It has virtually become conventional wisdom among practitioners in Indonesia that voters who initially support one particular candidate but don't receive gifts can end up voting for more cashed-up rivals. This is particularly so because although voters might be party loyalists, they still have to pick between several candidates from that party (see Chap. 1). As shown in the left panel of Fig. 4.5 above, 11% of surveyed candidates and 14% of brokers showered their supporters with benefits to protect their bases from other (internal) competitors. Hajriyanto Thohari from the nationalist Golkar party was told by his success team members that they needed to make cash payments in order to secure their votes a few days before the election. Hajriyanto mentioned a term being used to describe this practice: *tembakan terakhir* or 'final shot.' He said: "I rejected their suggestion. They replied that they wouldn't be responsible if I failed to get the seat. … I didn't make it, despite the fact that I had been previously elected many times in this electoral district" (Interview, 21 April 2014).

Similarly, a prominent PKB politician admitted: "I distributed patronage or money to tie them to me. … just small sums of money. … These minor gifts… ensure that the voters I had been cultivating don't switch to opponents" (Interview, 20 April 2014). As I will argue in more detail in Chap. 5, higher uncertainty as a result of the open ballot system and intense intraparty competition drives candidates to—in a term they frequently used— 'tie' (*mengikat*) their loyalists with cash, so they wouldn't turn their backs on them. These findings are somewhat similar to those of Dunning and Stokes (2008) in Mexico, where many initial supporters of the Party of the Institutionalised Revolution (PRI) finally voted for rivals because they didn't get any electoral incentives. In Indonesia, however, things are even more complicated because although voters might remain loyal to a party, they still can vote for a different candidate within that party. When parties must distribute benefits to sustain partisan loyalty, Dunning and Stokes called the relevant voters' attitude 'conditional loyalty.' Overall, despite the cultural norm of gift giving in Indonesia motivating candidates to cultivate their core supporters, the dominant discourse among candidates is that a significant proportion of their so-called loyalists would change their votes if they didn't receive assistance. Again, my findings from broker and candidate survey as well as qualitative interviews support the notion that partisan, loyalist voters are widely favoured as vote-buying targets among political actors.

#### *4.4.3 Evidence from Nationwide Surveys*

In this section, I return to analysing voter-level data in order to obtain more detailed information on who receives offers of material benefits, and when. By adding temporal dimensions these data (see Table 4.3 below) are again highly supportive of the party-loyalist strategy and explain in which period party loyalists receive the most offers of benefits. Non-party identifiers were consistently less likely to be offered gifts across four surveys in 2014 leading up to the election. Using the statistical analysis approach of relative risk or risk ratio (RR), the probability of self-proclaimed partisan voters experiencing vote buying was two or three times higher than that of non-identifiers. For example, in January 2014, 9.1% of the respondents with partisan closeness were being offered rewards, but only 3.5% of those with no partisanship were targeted—a proportional difference of (9.1 − 3.5)/3.5 or around 1.55 (or 155%). The same also held true in late February and early March 2014, about one and a half months before the election, with a proportional difference of 1.96 or 196%.

As the election approached, the pace of vote buying accelerated as both party loyalists and uncommitted voters were increasingly targeted. In April 2014, about 25% of respondents were exposed to such transactions. However, when restricting my analysis only to those with or without partisanship, 43.1% of the party identifiers were offered rewards. By now, 21.8% of the non-identifiers had been offered rewards too, a big jump from 3.5% in January, 6.5% in late February to early March, and 9.1% in the end of March. Once again, consider that this segment of non-identifiers constitutes 85% of the total electorate. Indeed, it appears that the machines were still taking care of core constituencies, with proportional differences of 1.18 (118%) and 0.98 (98%) in late March 2014 and April 2014, respectively. Relative to previous results, however, the overall percentage of investment in core voters slightly decreased, while that in non-partisan voters markedly increased. The statistical risk ratio of partisans being targeted in January and February–March 2014 was two to three times more likely than that of non-identifiers, but this probability slightly decreased to only two times in the end of March and April 2014 (Table 4.3).


**Table 4.3** Individual vote buying by some or no partisanship (%)

Source: The January, February–March, and late March 2014 data were taken from my pre-election surveys, while the April 2014 numbers were drawn from my post-election survey (see Appendix A)

From the data above, it is evident that candidates simultaneously target party loyalists and non-partisan voters, and the proportion of the latter increases as election day nears (while partisans are still clearly preferred in relative terms). However, such interpretations contradict the results from my sample survey of low-level politicians and brokers, as well as my in-depth interviews with high-level politicians. These political operators repeatedly expressed a strong preference for targeting loyalists. They believed that investment in uncommitted voters might be wasted. If they showed such strong preference to capture core supporters in their strategic thinking, why did they end up distributing so much cash and goods to so many non-partisan voters? This question is particularly relevant given that, it will be recalled, that in total numbers, non-party identifiers heavily outnumbered party identifiers, meaning that more voters without partisan attachments received offers than those with such attachments. This is the puzzle I develop in the last sections of this chapter—a puzzle that the following chapters then try to resolve.

### 4.5 Mixed Results

The puzzle described above is further illustrated in a last presentation of survey results comparing levels of vote buying in the 2014 and 2009 elections in relation to voters' partisanship. As shown in Figs. 4.6 and 4.7, the incidence of vote buying increased from 2009 to 2014, but the pattern in regards to partisanship remained the same: in relative terms, partisan

**Fig. 4.6** Gift receipt by partisan effect: Direct vote buying (%). Source: My postelection survey, 22–26 April 2014

**Fig. 4.7** Gift receipt by partisan effect: Neighbourhood vote buying (%). Source: My post-election survey, 22–26 April 2014

voters were more likely to be targeted, but in total numbers, uncommitted voters who received benefits outnumbered the core voters who did so.

If we use a purely percentage-based analysis, the analysis of the survey data, and of the Indonesian case, presented so far provides strong support for the core-voter model, according to which partisan orientation increases the likelihood of a person receiving benefits during an election. This is true for both direct (Fig. 4.6) and neighbourhood measures (Fig. 4.7), which consistently show that if a voter identified with a party, then he or she was more likely to receive offers of vote buying than someone who wasn't a partisan.

If we rely on the *absolute* approach, by contrast, we come to a different conclusion—one that the swing-voter school of vote buying would prefer. As shown in Fig. 4.6, 'only' 22% of non-partisans received gifts during the 2014 campaign. However, these are 22% of the vast majority of voters (85% of the total electorate). By contrast, the 43% of loyal, partisan voters who received gifts sounds large, but they only make up a tiny percentage of the electorate. Put in a different way, 22% of 85% is much larger than 43% of 15%. Recall that there were approximately 187 million voters in the 2014 election. This means there were almost 35 million non-partisan voters who received offers of vote buying but only 12 million partisan voters were exposed to such exchange. The numbers also bear out that vote buying increased faster among uncommitted voters than among party loyalists. It roughly tripled in the former category, while it only approximately doubled in the latter between 2009 and 2014 (based on the direct measure). Clearly, the voter survey data lead to differing—and, depending on the approach, even conflicting—interpretations on how to locate Indonesia in terms of the swing- versus loyalist-voter debate on vote-buying targets.

#### 4.6 Conclusion

This chapter has discussed—from a number of methodological, empirical, and statistical perspectives—the effect of partisanship on vote buying. For the purpose of this study, the finding that party identification levels in Indonesia are low is crucial—only 15% of the electorate feel close to a political party. In order to avoid vulnerability to endogeneity assumptions, I further demonstrated that partisanship—as measured by my surveys isn't endogenous to electoral bribes. From this firmer methodological platform, I was able to show that there are variations in the effects of partisanship across the party spectrum, and in terms of the receipt of votebuying offer. Two different statistical analyses allowed us to assess that relative to other partisanships, those who are aligned with PDI-P, Golkar, and the Democratic Party (and other smaller parties depending on the measure) have a higher probability of being targeted with benefits than partisans of other parties. Roughly, the larger the party, the higher the chance a loyalist to that party has of being hit by vote buying.

But given that partisans are low in number and non-partisans constitute the vast bulk of the electorate, how do political operators themselves explain how they direct their limited resources during campaigns? My novel dataset from low-level politicians and brokers found that they indeed professed that they targeted loyalists. This pattern was confirmed by in-depth interviews with high-level politicians. The rationales behind this preferred strategy were associated with risk aversion and the moral economy of vote buying. Yet additional individual-level data showed that while in relative terms, partisans were more likely to be targeted, in absolute terms, more swing voters received rewards. In essence, then, neither the core-voter nor the swing-voter model is fully applicable in Indonesia. Political operators claim to target loyalists, and they indeed reach a significant number of them with their vote-buying efforts. But vote buying is so extensive and the number of partisans so small, that in reality much of the cash and gifts are absorbed by swing voters. Is this the result of failed targeting on the part of the operators? Or did candidates and brokers misunderstand who exactly was a loyalist and who was a swing voter, leading them to believe they targeted the former but ultimately hit the latter? Or were other mechanisms at work? The discussion above systematically set up this puzzle, which I address in the next chapter.

#### Bibliography

Books, Journal Articles, and Unpublished Papers



**Open Access** This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

<sup>———. (2010)</sup> Politics and Poverty: Electoral Clientelism in Latin America. PhD thesis, University of California at Berkeley, USA.

# How Targeting Goes Astray: Explaining the Gap Between Intentions and Outcomes

This chapter addresses puzzles that were identified in the previous chapter. First, if partisan voters have a higher probability of being targeted for vote buying, while mass partisanship in Indonesia is comparatively weak, how feasible is it for candidates to win only by targeting partisans? Second, how can we explain the gap between politicians' insistence that they are predominantly targeting party loyalists and the reality that so much cash and goods end up in the hands of non-partisan voters? This second puzzle raises a further question about the criteria that politicians and brokers actually use when identifying the loyal voters they wish to target.

With respect to the first of these puzzles, I argue that given partisan voters are limited in number and fought over among multiple candidates from the same party, candidates first target the party constituencies they think will be likely to support them personally. However, to clinch victory, they need to seek support beyond their own traditional party bases, and they leverage personal connections in order to do so. This, in turn, has to do with the second puzzle: given they are so dependent on personal networks, candidates and brokers misrepresent personal connections as partisan leanings. They say they are targeting partisan voters, when in reality they are targeting voters who are connected to them, or to their brokers, by personal ties. By emphasising this point, I offer a contribution to the long-standing debate between advocates of core- versus swing-voter models of vote buying by integrating analysis of the core-voter strategy with an emphasis on personal networks. I argue that candidates and brokers' targeting strategies are misdirected; they plan to target partisan voters, but in fact they distribute most benefits to voters embedded in personal networks, who they presume to be loyal (but in fact are quite often only weakly connected to them). Put simply, candidates and brokers misidentify, or even exaggerate, the number of partisan voters because they typically blur the line between partisan and personal loyalties.

This interpretation, however, leaves a question unanswered: why should candidates get it so wrong? We know from comparative literature that brokers have an incentive to exaggerate the number of partisan, loyalist voters,1 even to deceive their candidates on this issue, so that they can engage in predation. But candidates don't have any such incentive. In addition to the mistargeting story, this chapter accordingly argues that many of the people who are connected through personal networks are in fact not even loyal to the candidate. This is largely because candidates are targeting people through other networks in which brokers may lack close personal ties to them, leading to agency loss. Finally, I argue that much of what I observe in Indonesia is like neither the core-voter nor the swingvoter model. This is largely because literature proposing these models was developed in settings that differ significantly from Indonesia. As a result, the dynamics of vote buying, or at least the targeting of vote buying, are distinctive in Indonesia.

#### 5.1 Contextual Factors

As discussed in Chap. 1, vote buying doesn't occur in isolation. It is therefore important to take into account its institutional context. Our first challenge is to assess whether Indonesia's context differs from those under which the swing versus loyalist debate arose. In shining a light on context, this chapter compares Indonesia to several Latin American countries on which the literature is based, in terms of five contextual factors: (1) organisation of vote buying, (2) electoral system, (3) party base, (4) level of partisanship, and (5) voting systems.

<sup>1</sup> In this study, the word 'partisan' is sometimes used interchangeably with 'loyalist.' However, I should stress that there is still a significant difference between the two. I define the first in party terms and the second in candidate terms. However, as I will argue later, given the nature of open-list PR which incentivises personal votes, candidates will 'personalise' those who are loyal to their party in the sense that they will try to convert, or combine, their support for the party into personal support for them as individual candidates, over their co-partisan rivals.

In terms of the structure of vote buying, much of the relevant literature is produced out of Latin America, which assumes party-based voting.2 In such cases, we should not only expect to see a high level of partisan voting, but the party machine will obviously be at the centre of political campaigns. Indonesia is virtually the opposite. This is important because both the swing- and core-voter models arose in settings where party machines have the capacity to monitor recipients' votes to ensure that bargains are kept. Political parties in Argentina are bottom-heavy organisations that embed themselves socially into communities, relying on an army of brokers linked to their local party structure (Stokes, 2007: 82–83; Kramon, 2013: 11). This organisation stems largely from the capacity of parties to control the distribution of state resources, enabling them to engage in deep interactions with voters (Auyero, 2001). The same holds true in other Latin American countries such as Mexico and Venezuela (Table 5.1). The assumption that typically underlies the literature on turnout buying (e.g. Nichter, 2008; Cox and McCubbins, 1986) is that parties are the only dominant actors in delivering benefits to their supporters. Even the norms of reciprocity model of Finan and Schechter (2012) require the existence of party operatives who can develop iterative relationships with voters and therefore channel rewards to reciprocal individuals. As Kramon (2013) pointed out, if political parties don't possess such personalised relationships with voters, vote buying is unlikely to be effective in encouraging a sense of moral obligation on the part of voters.

2Among others, Stokes (2005) provides evidence of this from Argentina. Enforcing votebuying bargains among swing or weakly opposed voters in Argentina, she argues, requires a well-organised party structure. Auyero (2001) and Levitsky (2003) also worked on Argentina, but focus on how party brokers function in everyday politics in the country. Similarly, in a later work, Stokes and her colleagues (2013) found evidence of the core-voter tendency in Argentina, Venezuela, and Mexico (as well as India), while emphasising the role of political parties as the main distributor of clientelist exchange. Similarly, Diaz-Cayeros and his colleagues (2012, 2016) found that political operatives in Mexico engage in more turnout buying (targeting passive loyalists to ensure they vote) rather than vote buying (favouring swing voters to sway their vote). The rationale is to avoid envy and jealousy among core constituents who might defect if not given benefits. Gans-Morse et al. (2014) generated valuable insights into the role of political context in Argentina and Brazil in shaping vote buying. In their model, given Brazil's strictly enforced compulsory voting and low machine support, party machines predominantly targeted swing voters rather than passive supporters. The point is that this literature on organisation of vote buying in Latin American countries focuses on political parties.


**Table 5.1** The different contexts of Indonesia and Latin American countries

Scholars of Indonesia have come to the conclusion that most parties in Indonesia are less organised as political machines, and they generally lack the capacity and organisational structure to penetrate local communities (Tomsa and Ufen, 2012; Berenschot, 2015). Compared to many countries in South America, the mobilisational strength of Indonesian parties is generally weak, and they are hardly involved in any kind of constituency service. In the conventional literature on vote buying, constituency service plays a significant role in increasing interactions between parties and voters (e.g. Auyero, 2001; Diaz-Cayeros et al., 2016). This is largely because, Berenschot (2015) argues, the distribution of state resources in Indonesia is largely not under the control of political parties, but instead in the bureaucracy, which has maintained significant discretionary power since the authoritarian period. As Berenschot (2015: 560) put it with regard to Latin America, "the degree of party control over the distribution of state resources seems to constitute an important contrast with Indonesia." This helps explain why most Indonesian parties are largely invisible at the grassroots and therefore lack the requisite capacity and organisational structure to engage with voters.

Put in a nutshell, the literature on the swing- and core-voter models arises in a context where party machines are both active in, and capable of, mobilising voters. Accordingly, the main organisational vehicle for vote buying in Latin America is parties. In Indonesia, by contrast, parties are largely inactive and inert, and it is individual candidates who run grassroots campaigns, including clientelist mobilisations.

Another difference with a lot of the Latin American cases on which the literature is based is the electoral system. In many countries in that region, MPs are elected through closed-list proportional representation and the remainder are elected by plurality rules (Table 5.1). Under the closed-list system, candidates rely primarily on party reputation, and competition primarily occurs as disputes between parties (Mainwaring, 1991). It therefore makes sense to target partisan voters and geographically discrete base areas to mobilise turnout. In this context, as Gans-Morse and his colleagues (2014: 417) put it, because parties and candidates have different electoral bases, they "don't directly compete to provide clientelist rewards to the *same* citizens" (italics in original).

Indonesia, by contrast, provides a completely different picture. Indonesia has adopted open-list elections. How does this difference make the logic of vote buying different in the Indonesian case? One obvious difference is that candidates need to rely on personal networks rather than the party. The open-list PR incentivises a personal vote and makes parties less relevant, which in turn lead elections to be more candidate-centric. Vote-seeking politicians wage highly personal campaigns, with little reference to their party platforms or policy positions. They also build personal campaign teams, independent of party structures. An overwhelming majority of candidates across the political spectrum, across distinct strata of legislative assemblies, rely on informal non-party organisations of brokers—known as *tim sukses* (success teams)—to mobilise voters. Since party structures are contested among other individuals in the party list, candidates cannot rely exclusively on them. This is obviously different from the context which has given rise to the dominant literature on vote buying or turnout buying in Latin American settings where political parties are well organised and socially embedded (e.g. Stokes, 2007; Diaz-Cayeros et al., 2016). While the literature conventionally assumes that brokers are "agents of a political party" (Holland and Palmer-Rubin, 2015: 1187), brokers in Indonesia are the soldiers of individual candidates. In Indonesia, candidates require their brokers to perform multiple tasks, ranging from constituency mobilisation, organising campaign rallies, distributing campaign paraphernalia, and delivering cash to voters.

Moreover, rather than winning the majority or large plurality of the votes that parties or candidates pursue in various Latin American countries, in Indonesia's open-list multi-member districts, a small proportion of the electorate can decide candidates' electoral fate. As already discussed in Chap. 1, given this system allows voters to determine which among a party's candidates are elected, it only requires candidates—provided they, their party, and their fellow party candidates have collectively won enough votes to gain one seat—to win just enough voters to beat their co-partisans. The marginal value of each voter collected through vote-buying strategies can be high enough to clinch victory. Nurul Arifin of Golkar colourfully talked of "the 2014 elections [being] like the civil war in Syria. Brothers are battling each other. The sword used in the battle is money to kill their own brothers" (*Merdeka.com*, 28 April 2014). She pointed her finger on the open-list PR's effect of making candidates of the same party fight each other. For Nurul, the only way to beat co-partisans is by outspending one's internal party rival in terms of distributing cash to voters; she claimed not to participate in vote buying and attributed her defeat to that fact (*Merdeka.com*, 28 April 2014).

Having discussed the first two distinguishing features of Indonesia's political environment relative to several Latin American countries, we now turn to reviewing the degree of partisan closeness in the two regions. Much of the debate on vote-buying strategies has arisen in contexts where levels of voter partisanship are higher than in Indonesia. Table 5.2 compares the level of partisanship in Indonesia with other countries (excluding Western countries) that have more or less adopted the Comparative Study of Electoral Systems (CSES) questions on partisanship. The most recent measure of mass partisanship in Indonesia conducted by my polling institute Indikator in February 2019 is 10.1%, far below the average level (49%).

With regard to the level of mass partisanship, Indonesia is different to Venezuela, Mexico, and Argentina and others, from which comparative literature on turnout buying has been produced, where levels of partisanship are relatively high. In such cases, the prospects of victory are much better for parties that simply target their partisan supporters with material inducements. This contrasts sharply with Indonesia where voters produce low scores on party identification.3 The country's condition of low party identification is similar to Brazil, where the recipients of largesse are predominantly swing voters (Nichter, 2010; Gans-Morse et al., 2014).

Figure 5.1 reveals the total percentage of Indonesians who reported feeling close to any party from 1999 to 2015. The percentage between July 1999 and November 2004 exceeded 50%, reflecting the decline of the early enthusiasm that accompanied the transition to democracy and Indonesia's first post-Suharto elections in June 1999, but there was a further sharp decline in mass partisanship from December 2004 to December

3For further discussion on the aggregate level of mass partisanship in Indonesia, see Chap. 4.


**Table 5.2** Party closeness across the worlda

Sources: Data from South and Central American countries are taken from the Latinobarometer 2013 (see http://www.latinobarometro.org/latOnline.jsp. Accessed 14 May 2016) while those from African countries are provided from the Afrobarometer Round 5 2011–2012 (see http://afrobarometer.org/ online-data-analysis/analyse-online. Accessed 14 May 2016). Data from some Asian countries are taken from the Asian Barometer 2 (2005–2008) (see http://www.jdsurvey.net/jds/jdsurveyAnalisis.jsp?ES\_ COL=101&Idioma=I&Seccion Col=06&ESID=503. Accessed 18 May 2016)

The question used in the Americas is "Is there any political party you feel closer to than others?" whereas in Africa it is: "Do you feel close to any particular political party?" The question used by the Asian Barometer is: "Among the political parties listed here, which party if any do you feel closest to?"

a One important note is that partisanship in Vietnam which reached up to 84% clearly means something different under one-party rule than party identification in (multiparty) democracies

2005, with the aggregate level of partisan allegiance decreasing significantly to less than 30%. Data from 2006 to 2010 exhibit increasingly weak party identification, with the level of party allegiance shrinking to an average of 22.2%. Since 2011, the degree of partisan affiliation of the electorate has fallen below 20%. This low level of partisanship isn't superficial but is consistent.

One contextual factor that helps explain the level of partisan loyalty is the political party system. This is another difference between Latin American cases and Indonesia. Political scientists have long argued that party identification tends to be lower in multiparty systems (e.g. Thomassen and Rosema, 2009: 50). In two-dominant party systems, it tends to be higher. Countries like Argentina, Venezuela, and, to a lesser extent, Mexico, where much of the scholarly literature has identified turnoutbuying strategies, have two (or three) dominant parties that have decadesold historical roots. Parties have strong 'base areas' in these countries. Take, for instance, working-class areas in Argentina that have been voting for Peronists for years, or the long-time rural voter base of the Party of the Institutionalized Revolution (PRI) in Mexico.

The Indonesia case, however, is completely different. There has been a highly fragmented political party system in the country during the periods of competitive democracy (1950s and post-1998). As a result, only a few political parties in Indonesia enjoy a strong tradition of having partisan voters. PDI-P is perhaps the strongest example because there are some areas, especially in Central Java and North Sumatra, where people have been voting for the party for years, and where they can go back further and trace a history of the Indonesian Nationalist Party (PNI) dominance in the 1950s. As thoroughly discussed in Chap. 1, PDI-P has taken root among less pious Muslims, middle-lower-income groups, and religious minorities. Another party that deserves to be mentioned is PKB, which has profited from close ties with the large socio-religious organisation, NU. It has consistently maintained its electoral base, especially within traditionalist communities in East Java. In general, however, most political parties in Indonesia don't have large or long-standing partisan bases.

Finally, much of this literature occurs in settings where voting is compulsory such as in Brazil, Argentina, and Mexico. Intuitively, compulsory voting increases the likelihood of parties targeting swing voters with payments, while the party loyalist strategy is more likely to develop in environments where voting in elections is optional. In places where voting is compulsory, parties logically will not fear non-participation by their own supporters because they may be subject to a penalty if they don't cast their votes. In such places, uncommitted swing voters who are willing to sell their votes are also more likely to participate. Empirical evidence, however, shows mixed results. Where compulsory voting laws are weakly enforced, as in Argentina, we can see evidence of turnout buying. In contrast, strict enforcement of compulsory voting encourages machines to target swing voters in Brazil (Gans-Morse et al., 2014; Nichter, 2010). In Indonesia, voting isn't compulsory. Under such circumstances, lukewarm supporters might not show up at the polls, if not given a benefit, making payments to such voters more likely. This gives rise to what Schaffer and Schedler (2007: 17–18) described as 'participation buying.'

Targeting party loyalists may help reduce uncertainty on the part of candidates regarding the final electoral results, as suggested in the preceding chapter. If a party loyalist bothers to turn out, the paying candidates can be reasonably sure that the voter will support them. However, it doesn't make this uncertainty completely disappear because the voter may, of course, choose a different candidate from the same party. Even if the party loyalist strategy is plausible in the context of non-compulsory voting, the emphasis on the personal vote under the open-list systems adopted by Indonesia means that candidates from the same party have to compete against each other for the support of a limited pool of partisan voters.

Overall, much of the scholarly literature analyses vote buying in the context of well-organised parties. It was also developed in a context where political parties are more prominent in organising election campaigns. Equally, the existing literature arose to describe environments where parties have some very strong 'base areas.' Indonesia, however, is different from these conditions. Vote-seeking politicians in Indonesia operate in a setting where the structure of vote buying is largely not party based, and political parties play, overall, a marginal role in election campaigns at the grassroots. The personalised nature of voting allows candidate to win only by gaining the support of a small slice of the electorate and in conditions where partisan ties are comparatively weak.

### 5.2 Solving the Puzzles

Having explained the contextual factors that affect the strategies candidates employ, we are now in a position to review the two puzzles stated at the outset of this chapter. In this section, I discuss the first puzzle about the feasibility for candidates to win by targeting only partisans, given their low absolute number and the strong intraparty competition among candidates for the support of such voters.

In an attempt to answer the first puzzle, I propose an explanation for how vote buying works in Indonesia that is distinctive from the swing- and core-voter models. I do so by drawing on the core-voter model, which emphasises turnout buying targeting party loyalists, but combine this analysis with a strong emphasis on personal networks. I argue that in Indonesia candidates and brokers actually intend to target partisan voters, but in reality, they mostly distribute patronage to people who are connected to them via personal networks. I call this combination of features a 'personal loyalist strategy.' Allowing a role for partisan voters, this strategy tries to personalise such voters by making them loyal not only to the party but also to the individual candidate within the party, as will be elaborated below.

The personal loyalist strategy, which combines targeting personalised party constituents with reliance on personal networks, is best suited to the context of an open-list PR system like Indonesia which provides incentives to politicians to rely on personal networks rather than the party, and in settings where the loyalties involved are largely personalised. The approach is built upon three distinct but interrelated strategies that candidates typically use. First of all, when they can, candidates start by targeting areas they think of as being 'party bases' on the basis of previous election returns. They believe these areas will likely yield greatest success in terms of personal votes. But, given there are only a few party strongholds, and these are typically contested among co-partisans, every candidate, starting long before the election, seeks to personalise their campaign in order to maximise their individual votes. They do so by reaching down to voters within this party base as individuals, including via party operatives whom they treat as personal clients.

Second, having realised that their party loyalists are limited in number and vulnerable to competition from internal competitors, candidates seek to expand their electoral base by using whatever personal connections they can mobilise. They start by determining their 'base' not only on the basis of previous electoral patterns but also based on their birthplace, kinship, religious, or ethnic ties, as I elaborate in more detail in the next section. In addition, candidates also choose areas where they have network connections or where they have in the past provided patronage and constituency service (Aspinall et al. 2017: 13). This is typically not enough, though. If they rely solely on their personal networks, candidates would fail to gain the votes they need to win. Recall that candidates running for nationallevel and provincial-level parliaments have to win hundreds of thousands of votes, depending on the density of the population.

Finally, in order to further enhance their electoral prospects, candidates seek to reach out to a greater number of voters through personal connections mediated by non-party brokers. These brokers expand beyond the party and the candidate's personal electoral base by providing additional support on the basis of their own clientelist and other personal networks. As Aspinall and his colleagues (2017: 5) nicely put it, "brokerage networks were a method for scaling up a candidate's personal patron-client networks to encompass a greater number of voters." Therefore, I argue that what matters here isn't the broker per se, but the *size* of the broker army who run vote-buying efforts on the basis of their heterogeneous networks. How large and strong the brokerage network a candidate has can determine whether that candidate wins more votes than his or her internal rivals.

At first sight, this personal loyalist strategy looks promising. With limited resources in hand and dealing with large constituencies, candidates strategically target only those they think will be most likely to support them personally. However, as presented in Chap. 7, most of the recipients simply take the money but vote according to their conscience. A senior journalist in Padang, West Sumatra, cited an analogy of hiring a person to climb coconut trees. He said giving money to voters is like when a coconut picker takes the salary, but declines to pick the coconuts (Sukri Umar, Interview 23 September 2014).4 Notwithstanding candidates' strong intentions to reward only people they think of as loyal supporters, this strategy cannot fully prevent 'leakage' or 'slippage' to people who don't repay them with support.5 This is largely because many of those candidates depicted as 'loyalists' are likely to defect, as will be explained in more detail in the following pages. The distinction between 'party base' and 'personal vote base' is thus blurred, as Aspinall and his colleagues (2017: 13) explain:

4This strategy isn't only unique in the Indonesian context. Cardinal Sin, Archbishop of Manila, for instance, is widely known for his advice by giving a famous analogy: "take the bait, but not the hook" (Schaffer, 2005; Hicken et al., 2017).

<sup>5</sup> It is striking that candidates keep thinking of such people as their 'base,' while they should know that a lot of them are only weakly connected to them. As I argue in Chap. 7, candidates actually understand that the targeting of vote buying is misdirected, but because their goal is only to obtain a narrow winning margin, they know that the strategy is still worthwhile.

Overall, when candidates used terms suggesting core versus swing voters, they thus typically understood such terms in personal terms, rather than in terms of voters' identification with a party or program. They were talking about a clientele rather than a party core.

This brings us back to the seemingly paradoxical finding in my voter survey described in Chap. 4: while partisan voters are more likely to be targeted than non-partisans, in fact, most vote buying happens among uncommitted voters. This outcome, I argue, is a result of weaknesses of the strategy above that combines a party loyalist approach with reliance on personal networks. But why is the conceptual confusion? Why do candidates keep insisting they are targeting loyal voters, while the facts show that much of their spending is wasted on uncommitted voters who receive benefits but don't always reciprocate with their votes? Three preliminary explanations are offered at this stage: (1) candidates and brokers tend to exaggerate the number of partisan voters; (2) they exhibit confusion about the concept of 'loyalty'; and (3) agency loss occurs, contributing to a large amount of targeting of uncommitted voters.

#### *5.2.1 Exaggerating the Numbers of Partisan, Loyalist Voters*

The first probable explanation for the gap is clear: many politicians across the political spectrum exaggerate the number of partisan voters in Indonesia. Many politicians find it hard to accept the reality that relatively few voters feel close to any political party in Indonesia today. As one of the leading public commentators on Indonesian political affairs, with regular national media exposure due in part to my position with the country's pre-eminent opinion survey organisations, LSI and Indikator, many political parties frequently invite me to discuss political trends and, especially, to explain survey data. When I present on the general decline in partisanship, party leaders often admit there is such a trend, but stubbornly reject the finding that as many as nine out of ten Indonesians express that they aren't close to any political party. They typically claim that the level of partisan loyalty isn't that low. Politicians not only express this denial in internal and closed discussions but also in public forums. In the lead-up to the 2014 election, I appeared on a prime-time television political programme with a party leader from the Islamist party, PPP. He expressed his strong disagreement with my survey findings, suggesting low scores on voters' psychological bonds with political parties. Similarly, in a confidential interview, one elected candidate from the traditionalist Islam party PKB suspected that many pollsters have consistently underestimated how many people feel close to political parties, especially his own party, by arguing:

It is difficult to believe that people who self-describe as being PKB loyalists are much fewer than those of the PKS. My party is closely associated with the largest Islamic organisation and it can be traced back in the history of NU dominance in Java during the 1955 elections…. So weird. (Interview, 12 May 2014)

The tendency to exaggerate the number of partisan voters might help increase politicians' confidence in the context of competitive elections and ameliorate their anxiety about future election results, as discussed below. However, although many politicians exaggerate the number of partisan voters, overall, they typically agree that more Indonesians match the nonpartisan profile than are partisans. They understand that many voters don't have partisan attachments, but insist a significant proportion of the electorate still feel close to a party.

Another powerful example of exaggeration is provided by a unique set of polling data from low-level politicians and brokers. As previously explained, from September to October 2014, I surveyed 299 candidates and 900 brokers in four provinces (Appendix B). Figure 5.2 shows that, overall, respondents tended to overestimate the number of partisan, loyalist voters. The prompt is: "We would like to inquire further about the characteristics of voters in your area. Out of ten people in your area,6 approximately how many people who always vote in every election and would always vote for the party/candidate7 you support?" Hence, voter loyalty here is defined in terms of turnout propensities and partisan closeness or loyalty to the candidate. While responses to this question were quite scattered, the modal answer—from either the surveyed candidates (featured in horizontal stripes within Fig. 5.2) or brokers (highlighted in vertical stripes)—was '7 out of 10.' No candidates said 'none,' whereas about 74% of candidates estimated the frequency of loyal voters in their regions at six out of ten or higher. Similarly, the overwhelming majority of brokers (74%) claimed that of ten people in

<sup>6</sup> In my questionnaire, I used undefined 'wilayah' (area) which didn't specifically refer to electoral district or village or neighbourhood.

7Having used the wording 'the party/candidate,' the question problematically conflated partisanship with a sense of loyalty to the candidate. Unfortunately, this part of the survey was designed before I decided to focus on the issue of partisanship in my follow-up research.

**Fig. 5.2** Political actors' perceptions of the frequency of loyal voters (%). The question used is "In your region, out of ten typical voters how many people would who always vote in every election and would always vote for the party/candidate you support?" Source: My survey of low-level politicians and brokers, 30 September–25 October 2014

their neighbourhoods, six to all could be categorised as certain loyal voters. Only 23% of brokers estimated the frequency of perceived loyalists at one to five out of ten people in their neighbourhoods. This response is, of course, a huge exaggeration, though presumably some bravado was involved.

Before we move on, it is worth pausing to consider one more source of information to check whether candidates and brokers really are exaggerating the number of partisan voters (so far, I have only relied on self-reported levels of partisanship in surveys). One way to think about party loyalist voters is that they are those voters who faithfully return to their party at every election. Accordingly, I provide additional data by estimating the number of partisan voters based on those who supported a party in the 2009 legislative election and voted for the same party in 2014.8 It must be

8The question reads: "Did you vote in the 2009 elections?" If the respondent responded affirmatively, we asked "Which party did you vote for?" The interviewer showed the list of political parties competing in the 2009 election. In the survey, we also asked: "Did you vote in the 2014 elections?" If yes, we asked "Which party did you vote for in the 2014 elections?" The interviewer showed the list of political parties competing in the 2014 election to the respondents.

noted, however, that the number of national parties competing in these elections didn't remain the same (38 parties contested in 2009 compared with 12 parties in 2014). It means my estimate is only able to measure the loyalty of voters whose parties competed in both 2009 and 2014. This calculation cannot also exclude 2014 first-time voters who were, of course, ineligible to vote in 2009.

My strategy is simple: drawing from my post-election survey in 2014 (see Appendix A), I ran cross-tabulation between those who reported their party choices in 2009 and 2014 to get the detailed percentages of party loyalists and non-loyalists in each party. As shown in Table 5.3 (look especially at the numbers without brackets), the proportion of party loyalists, by this measure, varies considerably. Only 20.7% of voters who sup-


**Table 5.3** Estimated numbers of party loyalists in Indonesia measured by voters who voted for the same party in two consecutive elections, 2009 and 2014

Source: The data on those who supported a party in the 2009 legislative election and voted for the same party in 2014 are collected from my post-election survey, 22–26 April 2014, while the data on the 2009 election results are taken from the Indonesia's Electoral Commission (Komisi Pemilihan Umum, KPU)

a The first figure in each of the last two columns is an estimate based on the total number of valid votes; the second number is based on the DPT

ported the Democratic Party in 2009 voted for the party again in 2014. This is largely because the party's popularity dropped significantly in 2014, following a series of party corruption scandals (Aspinall et al., 2015). Parties that were relatively successful at maintaining their voters' loyalty were Hanura, PDI-P, Gerindra, and Golkar. Note that the sampling frame of my voter-level data isn't only limited to the total number of valid votes casted in the election but all registered voters including those who didn't turn out on voting day. Accordingly, to produce an estimate of the proportion of eligible voters who are loyal to a particular party, I also convert these percentages of party loyalists and non-loyalists in each party into percentages of the fixed voter list or *Daftar Pemilih Tetap—*DPT (look for the number in brackets). I prefer to divide by the DPT rather than by the total number of valid votes because these data are based on the postelection survey whose target population weren't only those who cast their votes on election day but also those who didn't turn out.

Though using this rather weak measure of party loyalty (voting for the same party in two consecutive elections), the conclusion still holds that the number of party loyalists in Indonesia is relatively low. Only 22.2% of the total electorate voted in 2014 for the same party they had supported in 2009. Of votes cast, 36.6% were cast by people voting for the party they had supported in 2009. There isn't a huge difference between these figures, especially the first one, and the number of voters who reported themselves as being close to a party (15%) shortly after the 2014 election isn't significant, implying that the estimated proportion of party loyalists in Indonesia lay somewhere between 15% and 22% at that time. If we could compile data about voter loyalty based on those who faithfully return to their party in every election (i.e. 1999, 2004, 2009, and 2014), I suspect the percentage of party loyalists would be significantly lower than 22.2%. In short, a range of between 15% and 22% of party loyalists provides strong evidence that politicians and brokers make exaggerated claims about the number of such voters.

#### *5.2.2 The Amorphous Concept of Loyalty*

A second reasonable explanation for why so much cash ends up in the hands of voters who aren't especially attached to any particular party, despite strong desires of politicians and brokers to target loyal voters, is the obscurity of the concepts of voter loyalty and base areas in the Indonesian context. Though they make bold claims about how many voters are loyal, in fact, many politicians and brokers use amorphous criteria in defining loyal and non-loyal voters.

Using a hypothetical scenario which is developed from Stokes and her collaborators (2013), this study asked the surveyed candidates and brokers about the effect of voter types (defined by turnout propensities, partisan affiliation, and loyalty to the candidate) on attracting benefits. The sum of their responses to this question is then classified based on their political parties. Table 5.4 shows that I found surprisingly little variation across


**Table 5.4** Variations across political parties in targeting strategies

Source: My survey of low-level politicians and brokers, 30 September–25 October 2014

The question used is: "Imagine this situation. There is a candidate together with his/her campaign team who offered 10 social assistance packages to mobilise voters. In reality, the candidate's success team member had 40 neighbours in need of social assistance. According to you, which group of voters will the campaign team direct more assistance to?"

political parties in the targeting strategies. Almost all candidates from different political parties, except PKPI, said they would prioritise 'loyal voters.' In fact, more than 80% of respondents said this in most cases.

But one interesting point is that the figure for the PDI-P isn't that high, despite that party actually having a strong party base. This corresponds with the narrative from PDI-P candidates that their support base is truly loyal and solid so they don't need to seduce them with payments to attend polling places. Ahmad Basarah, a successful PDI-P candidate, for instance, claimed that he didn't distribute money within his own party base and only paid for snacks, drinks, and cigarettes for the many informal meetings he held with his loyalists (Interview, 21 April 2014). This claim, however, contradicts what we found from voter surveys. As discussed in Chap. 4, those who felt close to PDI-P were significantly more likely than the average voter to be exposed to vote buying. However, the general rule remains the same: candidates and brokers tended to say they focus on mobilising their own party bases. This is particularly the case when they were asked about their strategies in the 2014 election. When we divided locales into three categories: party base, opposition base, and swing districts, their first priority was consistently to exploit the 'party base' (Table 5.5).


**Table 5.5** In 2014, where did you distribute largess to get votes? (%)

As shown in Table 5.5, candidates from PKS and NasDem and their political operatives tended to be more reluctant to report their targeting strategies in 2014. The same holds true in the case of PDI-P candidates, showing that 'only' 39.69% of the respondents from the party's candidates and their brokers reported targeting their own base, while around a third of them were unwilling to respond to the question. However, the big picture is consistent: the so-called party base is the preferred target.9 Most politicians and their brokers prioritised their party bases because they thought their potential vote was higher there.

This is quite a striking result, not least because Tables 5.4 and 5.5 highlight evidence of targeting partisan voters in parties that don't really have a tradition of having partisan loyalist voters. As noted earlier, Indonesia has very few political parties which have enjoyed consistent political support across several elections. It is especially striking that candidates from NasDem claimed to be targeting their own loyal supporters, while in fact the party had just been officially launched in 2011 and was running for the first time in 2014. Hanura and Gerindra were quite similar. They competed for the first time in 2009 and their shares of the vote were quite minimal at the time. Accordingly, these three parties must have had only very small numbers of truly 'loyal' partisan voters.

Note that in the lead-up to the 2014 election, partisan voters constituted only 15% of the total electorate (see Chap. 4). Even if we use another criterion of voter loyalty based on those who voted for the same party in 2014 and 2009, we have only around 22% of the total electorate. These small numbers of party loyalists, to be sure, are highly contested, with internal party rivals competing for their support in the context of an open-list PR system which intensifies intense intraparty contests.

There is a widely held assumption that candidates who are placed high on party lists tend to rely on party members in their success teams because they are usually heads of party branches or close to key party leaders. Intuitively, these candidates will also tend to target party loyalists more than they target unattached voters. Table 5.6 confirms this practitioner's

<sup>9</sup> Interestingly, candidates and brokers from Central Java and North Sulawesi were more likely to target party loyalists than those from West Sumatra and East Java. This is perhaps related to the evidence that partisan voters are more likely to be found in Central Java and North Sulawesi (see Chaps. 1 and 4).


**Table 5.6** Candidates' position on the party list and targeting strategies (%)

Note: The 'refuse to answer' and 'don't know' options were excluded from the analysis

rule of thumb. The bivariate test with Pearson's chi-square shows that the relationship between candidates' position on the party list and targeting strategies reaches the 0.05 level of statistical significance.

#### *5.2.3 Agency Loss*

In addition to the two above factors, the problem of targeting gets worse as a result of agency loss. As discussed earlier, candidates and brokers tend to inflate the number of partisan voters because they actually conflate those who are loyal to the party and those who are perceived to be personally loyal to the candidate. However, this isn't the only problem. Brokers clearly have an interest to overstate their capacity to sway people to support their party/candidate. Brokers frequently try to convince their candidates they have leverage in a particular subdistrict or village, within an ethnic organisation, or inside some other social network (Warburton, 2016: 343), but in fact their claims are often baseless. That is similar to the point made by Stokes et al. (2013), who argue that brokers have an incentive in exaggerating their followers in order to extract rents. They argue that party leaders tend to favour targeting swing voters, but they are unable to control their party brokers who instead direct benefits to core supporters for rent extraction purposes. Such difficulty in controlling and monitoring brokers can be a source of substantial 'leakages' which can help explain why many supposedly 'loyal' voters targeted by machines in fact don't reciprocate with their votes: in the context of an overall turnoutbuying strategy brokers might (reversing the story suggested by Stokes et al., 2013) exaggerate the number of loyal voters to ensure a greater amount of money and goods flow through their hands.

But why do candidates also exaggerate? It is reasonable to assume that candidates cannot be easily fooled by their brokers. As we shall see, there is plenty of evidence that candidates do know that a lot of people who are identified through personal networks, including those who are provided by their brokers, are only weakly connected to them. They also fear that such voters aren't immune to approaches by brokers for rival candidates, and they generally also worry that their brokers might trick them for rent extraction purposes. Even so, vote buying remains an attractive investment for most candidates. We shall see in Chap. 7 that part of the explanation is that they still believe that there are a sufficient proportion of the electorate whose voting decisions are largely influenced by handouts, and this is more than enough to constitute narrow winning margins. Given the ballot secrecy that is strongly enforced in Indonesia (see Chap. 6), many candidates, especially those who won a seat, seemed to be lenient if their brokers didn't deliver all the votes they promised as long as they were successful in securing enough personal votes to win. Most candidates are still convinced that among those being targeted by their brokers, a proportion of recipients will reciprocate with votes. One prominent national politician from PKB, for instance, was quite relaxed about his broker performance. Despite handing out a large sum of money in 450,000 envelopes containing cash, he yielded 'only' around 123,000 votes and still got elected. He explained:

There is a consensus among candidates: If you only get a third of the total envelopes you distribute, that's a good result. First, set your target. If you want to receive 100,000 votes to get elected, you must distribute envelopes tripling that. That's the rule. (Interview, 20 April 2014)

Needless to say, many losing candidates still get quite upset if their brokers fall far short of their targets. Regardless of the outcome, most candidates actually realise that in the context of open-list PR systems, they rely heavily on brokers to reach out voters in search of personal votes and so "stand out in a crowded field of co-partisans" (Shugart, 2001: 183). As discussed in Chap. 6, when candidates started to exhaust the supply of partisans or those who have close personal ties to them, they often rely on their brokers to determine which voters should be targeted. These people who are selected through personal networks are loosely connected to ideological proximity to the party and may lack personal relationships with the candidate. Accordingly, the potential for leakage—defined as those who receive payment yet don't reciprocate with votes—is great. However, in a context of competitive elections like Indonesia where minor changes in voter support can make a difference to the outcome (see Chap. 7) and multiple candidates engage in vote buying, anxious candidates have little option but to rely on brokers to win the election. They need these brokers to expand their voter base.

In addition, candidates typically acknowledge that it is almost impossible to fully eliminate broker predation and to enforce voter compliance. Candidates are generally aware that to make vote-buying work, they need skilled and reliable brokers to distribute rewards to the voters they believe most likely to support them in return. In general terms, to do so, candidates would require a double-layered control mechanism to monitor both brokers and voters. The first layer of the mechanism aims primarily to discipline brokers in order to make vote buying more efficient and reduce incentives for broker predation. The second layer is devoted to dealing with the problem of voter compliance. In order to help detect who has kept with the bargain and enforced the deal, politicians require what Stokes (2005: 322) called "tentacle like organisational structures." Yet, in practice, most candidates in Indonesia fail to build impressive monitoring structures (Chap. 6). Equally, brokers have weak mechanisms to monitor their voters to ensure that they vote for the candidate. Accordingly, despite such a strong desire among candidates and brokers to appeal to constituents who they think will reciprocate, they mostly end up distributing resources to uncommitted voters. As the quotation from the PKB politician above suggests, they learn to live with considerable mistargeting and wastage.

To sum up the discussion, political machines plausibly exaggerate the proportion of loyalist voters. This is partly because the concept of loyal voters in the Indonesian context is ambiguous and obscure. When candidates and brokers routinely talk about targeting loyalist voters, they not only refer to fixed party loyalties but also judge loyalty in terms of personal networks.10 This confusion leads to misdirection of targeting strategies. Instead of directing campaign resources towards constituents who are truly loyal, most brokers end up being distributed to unattached voters, who don't necessarily reciprocate with support. Moreover, agency loss also produces the unreliability of voters who are identified through personal networks. These factors in combination—exaggerated claims about the number of loyalists, confusion of personal connections with loyalty, and poor voter compliance with clientelist deals—contribute to the large amount of targeting of uncommitted voters.

### 5.3 Dimensions of Personal Networks

Having understood the reasons why targeting goes astray, the subsequent question is: what sort of person does get targeted for the payments? What criteria are used in targeting those whom the candidates and brokers claim to be 'loyalists'? In seeking an answer, one quickly gets bogged down in a classic anthropological conundrum: the responses candidates and brokers provide in survey interviews may not be honest answers, or at least not responses that satisfy the observer (Eder, 1991: 153). But we can seek a satisfactory explanation from qualitative research. From my field interviews, it was obvious that when candidates claimed to be targeting 'loyalists,' they typically referred to more personalised rather than strictly party-based relationships.

As discussed in an earlier section of this chapter, context really matters in explaining this ambiguity. Given the nature of the non-party organisation of vote buying in Indonesia, candidates and brokers appear to extend the definition of loyal voters to include persons who are 'close' to them by virtue of brokerage networks, kinship, and patronage loyalties and other informal connections. In interviews, candidates and brokers always talked

10The point that, in practice, partisan and personal networks might be overlapping can be evident when political parties often recruit their supporters through personal or familial relationships. When a PDI-P candidate recruits a broker, this broker is likely to be from similar ideological and organisational networks as the candidate. Similarly, a PKB candidate typically mobilises brokers from the 'traditionalist' Islamic community in which the party was rooted. Then, the brokers are expected to target voters on the basis of nearness to their own personal networks. In other words, despite the salience of candidates' personal networks, to some extent, these networks should be situated within ideological, social-cultural, and religious milieus linked to their parties.

about targeting 'loyal voters' and 'base areas,' superficially sounding as though they were concerned with partisan loyalty and past voting patterns. However, when I pushed them and asked them what their 'basis' really was, it generally turned out to be about clientelism.11 For example, having a strong brokerage network in a particular village makes that area part of the 'base' of that candidate (Aspinall et al., 2017: 13). This is why, despite candidates and brokers claiming to be targeting loyal voters and base areas, they aren't truly pursuing a core-voter strategy as typically understood by scholars writing on other countries. Put differently, when candidates asked their brokers to exploit 'base areas,' they typically understand this phrase in personal rather than partisan terms (Aspinall et al., 2017: 13–14).

This is the main argument in the personal loyalist strategic logic I outline above. Candidates tend to be confused, and they wrongly think of personal loyalists as party partisans. The confusion over the definition of loyal voters provides suggestive evidence of the plausibility of my argument. Given the significance of personal connections in explaining targeting strategies, it is essential to better understand the various dimensions of personal networks in the Indonesian context.

One important foundation of personal networks in Indonesian electoral politics is areas where a candidate has close personal or familial connections. This typically includes the location of a candidate's birthplace, or where they grew up or lived, but can include other areas where they have family or marriage ties (e.g. a spouse's home village/subdistrict) (see Aspinall et al., 2017). When deciding on where to focus their campaign strategies, candidates generally draw in not only their immediate family but also to the greater family by using "bilateral kinship links from both parental lines and following the various branches of the family tree as far as they went" (Sumampouw, 2016: 326). For instance, Anton Miharjo who worked for Lathifah Shohib (a PKB candidate) told me a story. Lathifah is a distant cousin of the late Abdurrahman Wahid the former President of Indonesia and the grandson of the founder of NU, KH Hasyim Asy'ari. Lathifah who lived in Malang was a simple teacher and unable to mobilise a large campaign because of her limited finances. To

<sup>11</sup>Aspinall and his colleagues (2017) also had similar experiences when inquiring about candidates' targeting strategies. As Aspinall et al. (2017: 12) put it, most candidates claimed "they prioritized voters they considered to be their basis … believing this would yield greater return on their efforts."

attract loyalty from PKB and NU cadres, she hired a number of buses to travel to Wahid's tomb in Jombang and asked those who accompanied her from Malang to take the *baiat* or oath of allegiance to help win her campaign. Indeed, she finally got elected. Her campaign team organiser said to me, "Aside from the living, we also make use of the dead" (Interview, Anton Miharjo, 2 May 2014).

In ethnically and religiously divided areas, personal bases and networks can be also created on the basis of primordial affiliations. Ethnic associations in North Sumatra, for instance, played an important role in mobilising electoral support for candidates. Religious networks are equally important for many candidates. Candidates from the Batak ethnic group, for example, usually take advantage of *marga* (clan) networks to reach voters from the same clan (Interview, Firman Jaya Daeli, 21 July 2014). The same holds true in the case of many parts of eastern Indonesia where *fam*, derived from the Dutch *familienaam*, and equivalent to *marga*, is used to mobilise personal loyalty for candidates who share family connections (Sumampouw, 2016: 325). Further, in North Sulawesi, Christian candidates mostly targeted the Gereja Masehi Injili di Minahasa (GMIM, Christian Evangelical Church in Minahasa), the largest Protestant denomination in the province, and built personal associations with religious networks.

Personal networks can also refer to a group of voters having been previously targeted for patronage or constituency services. Viva Yoga Mauladi, a successful candidate from PAN, for instance, targeted what he called base areas where he had distributed patronage in the past (see also Aspinall et al., 2017). As a part of the leadership of a parliamentary commission for agriculture, plantations, forestry and maritime affairs, and fisheries and food affairs, he mostly funnelled his pork-barrel projects to villages or community groups in his electoral district where he had personal or network connections. Such practices weren't limited only to incumbent candidates who could access regular slush funds allocated to their constituents but were also practiced by non-incumbent yet well-resourced candidates who were widely known for their generosity in particular villages. Accordingly, as suggested by Aspinall and his colleagues (2017: 13), most candidates viewed loyal supporters in terms of networks rather than partisan and geographic terms. Viva Yoga, for example, who regularly distributed small-scale infrastructure to local farmer associations, plantation workers, or fishermen groups that were closely related to his portfolio, claimed that the members of the groups were his core constituencies (Interview, 22 April 2014).

Overall, the relationships between politicians, brokers, and voters have increasingly been less ideological and more about personal connections and trust. This helps explain why party switching or political turncoatism is pervasive in Indonesia, especially at the constituency level. Some district candidates in a number of regions in South Sumatra move to other parties before an election, but still get elected (Hilmin, interview, 20 November 2014).12 It is often the case that when they switch to a new party, they bring their brokerage networks with them. Most brokers simply follow their bosses because they are strongly motivated by personal ties with, and zeal regarding, the candidate rather than party links. If such a shift occurs, the brokers ask their voters to follow as well (Hilmin, interview, 20 November 2014). This couldn't occur were those brokers and their followers strongly tied to their parties. As previously discussed, the nature of vote buying in Indonesia is non-party based, so that brokers' commitment to the candidate and brokers' relationships with voters isn't a matter of ideological conviction about the candidate's party. Rather, it involves a clientelist relationship and personalised networks between candidates and brokers, and between brokers and voters.

#### 5.4 Maintaining 'Loyalty'

The discussion thus far has argued that personal networks matter most in explaining targeting strategies. This doesn't mean that candidates ignore areas which have traditionally been viewed as their party bases. As noted above, the personalised nature of voting under the open-list system encourages candidates to personalise party strongholds when pursuing personal votes against co-partisans. But in order to win, they also try to expand their electoral support by targeting areas where they potentially have personal bases mediated by their brokers. In this section, we will explore the next stage, which is when candidates start building intimate social relations with their so-called loyal supporters, trying to lock in their relationship through the provision benefits.

Hence, in order to build and maintain voter loyalty, patronage plays a crucial role as political glue that links voters to candidates and secures their

<sup>12</sup>For more discussion about party switching in the case of Indonesia, see Nathan Allen (2012).

votes. In fact, given the significance of patronage in binding voter loyalty, it is sometimes difficult to disentangle patronage from other dimensions of personal networks when explaining electoral outcomes (Warburton, 2016: 351). The surveyed candidates and brokers believed that 51% of their supposedly loyal voters had received 'a lot' or 'quite a lot' of assistance, including money or gifts or construction of roads, houses of worship, insurance, and the like, from the candidate. Meanwhile, only 21% believed their 'loyalists' hadn't received any inducements. This distribution of material benefits not only occurred during campaigns, most incumbent candidates poured rewards into their voter bases long before the election.13 Most candidates I encountered used a variety of terms for this practice, such as '*menyantuni*' or '*melayani*' (serve or provide services).14 A commonly shared view among candidates is that they distributed largesse to their constituents because these were the masses of whom they were sure rather than being indifferent groups whose voting behaviour couldn't be determined (see Chap. 4).

By targeting voters whom they suspected to be leaning to them, candidates could minimise potential wastage of their resources. Although targeting such voters wouldn't guarantee that such offers would always be reciprocated, at least candidates would feel more confident that their investments would be translated into more votes than if they distributed to those outside their personal networks of relationships. For instance, one campaign manager, affiliated with a Jakarta-based political consultancy firm, gave me an image of a truck filled with Muslim prayer clothing transported to a group of voters in one district in East Java, because they had committed to support a candidate he was working for (Interview, 2 May 2014). In sum, candidates and brokers preferentially target those who they think will be loyal and reward them with benefits.

But how sure were the candidates and brokers of the loyalty of these allegedly loyal voters? Figure 5.3 suggests that they actually believed the so-called loyal voters also operated according to a 'transactional' logic and were concerned about concrete benefits and immediate rewards. The modal answer to the question about the percentage of 'loyal' voters who

<sup>13</sup> Such ongoing relationships that involve reiterated exchanges largely happen among serious candidates—especially incumbents who had been cultivating their constituents long before the election. Many candidates—especially first-timers, however, have no opportunity to build ongoing relationships with voters partly because of their limited access to state patronage.

14Also, see Eve Warburton on the case in Southeast Sulawesi (2016: 358–359).

**Fig. 5.3** Loyalists who would change their vote if they stopped receiving assistance (%). Source: My survey of low-level politicians and brokers, 30 September–25 October 2014

would change their vote if they stopped receiving campaign largesse was 'quite a lot,' while about 39% of the sample saying 'a lot' or 'not a lot.' It is especially striking that—in the respondents' view—only 12% of respondents believed that all of their 'loyal' voters would remain loyal and would be willing to turn out to vote for them, even if they didn't receive any benefit.

Again, contextual factors really matter in explaining this finding. In particular, Indonesia's highly competitive electoral settings drive candidates to ensure that voters they cultivate will not betray them and switch to cashed-up rivals. The majority of political actors surveyed thought that a significant fraction of their so-called loyal supporters would change their loyalties if not given benefits. As I will discuss in Chap. 7, the high degree of uncertainty induced by zero-sum competition inside party lists encourages candidates to be more responsive to voter demands (Hobolt and Klemmensen, 2008). My finding confirms ethnographic work by Aspinall and his colleagues (2017) who found deep anxiety among politicians regarding their chances of winning and widespread doubt about the allegiance of perceived loyalists who weren't given payments. This is particularly the case when multiple candidates engage in vote buying, and they need to make cash payments—borrowing the words of a Golkar's candidate—'to secure' (*mengamankan*) their voter base (Interview, 21 April 2014).

This finding might raise doubts about causation. Do candidate consider these people to be loyal voters merely because they have received benefits from them? Or, does their loyalty come first, but need to be 'locked in' by gifts? Although the answer may not be quite clear-cut, I would argue that, at first, candidates map out areas or groups of voters whom they think of as sympathetic, or of having the potential to support them, on the basis of personal relationships, social networks, or past personal patronage. My field observations suggest that candidates then don't take the loyalty of such perceived supporters for granted, with the exception of those who have relational networks or a high degree of personal closeness to the candidates. Instead they use material benefits to lock in their 'loyalty.' The result is that it seems that a lot of what candidates and brokers mean by their 'loyal' voters is simply people or community groups they have provided with benefits. However, as suggested by Aristo Munandar, a Golkar candidate running for a provincial seat legislature in West Sumatra, these benefits don't always come from candidates' own purse, but they can be accessed from state resources. Aristo was proud to claim that, when he was the head of Agam district, he frequently made visits to his constituents, providing them benefits from state funds, often in the form of projects that allowed him to generate loyalty among recipients (Interview, 23 September 2014). Incumbent candidates often rely on public resources and direct them to their base areas as a means of building and maintaining voter loyalty. Then, it is almost as if there is an underlying cultural assumption: "I have helped you, so I can expect your help in return."

The next interesting question is this: if the loyalty of the core supporters depends so much on benefits being offered, what sort of language accompanies such transactions? In most cases, candidates draw upon the language of gift giving, labelling their gifts in religious terminology such as alms (*sedekah*), or other forms of charitable donations in Islam when attempting to engage the support of their constituents. At the point of exchange with all groups of the electorate, whether loyal or undecided voters, as observed closely by Aspinall and his collaborators (2017: 5) in Java, "candidates and brokers downplayed the significance of their cash gifts and emphasised their emotional bonds with recipients, in line with what we might expect from the anthropological literature."

However, I found a slight difference during my observations. When dealing especially with so-called uncommitted voters, whom political

**Fig. 5.4** Payments as a binding transaction or gift? Source: My survey of brokers, 30 September–25 October 2014

machines usually called the 'floating masses' (*massa mengambang*), candidates frequently used economic terms such as 'electoral investment,' 'electoral market price,' and so on. Needless to say, these conversations about the transactional character of floating masses take place behind the backs of the voters. To be fair, the negative characterisation isn't exclusively directed at such voters, suggesting that supposedly loyal voters are occasionally associated with a transnational logic too, if not secured with a gift. But candidates and brokers appeared to be more lenient or benevolent if the payments were directed to the so-called loyalists rather than uncommitted voters.

This qualitative finding corresponds with the results from my broker survey. As shown in Fig. 5.4, when asked what motivated candidates to engage in vote buying,15 while responses were quite scattered, a significant proportion of the sample said 'as a form of kickback to voters' (39%) and 'as compensation in lieu of wages lost because the voter had to leave work to come to the polls' (38%). About 26% of the respondents viewed the offer as 'a sign of gratitude or reciprocation from candidates.' Interestingly, when I ran cross-tabulation tables, brokers who targeted undecided voters were more likely to consider the gifts as 'a form of kickback to voters.' In

<sup>15</sup>The question reads: "According to you, what is the motivation of the candidate who distributed the envelope/money/goods to the voters prior to the election?" [answer can be more than one].

contrast, those brokers who were targeting loyal supporters were more likely to view the payment 'in lieu of income lost because voters had to leave work,' or 'as a sign of gratitude from the candidate,' or 'as a sign that the candidate is socially minded.' There are indications that intermediaries can tolerate such compensation for 'loyal' voters partly because they view such voters as being more reciprocal than undecided voters. Regardless of the fact that some perceived loyalists can end up voting for rivals, they were seen to be more reliable in delivering support to the distributing candidate.

Having presented the results of my study, one might question the meaning of the very concept of 'loyal voters' in the Indonesian context. If these voters are truly loyal, as claimed by candidates and brokers, why do they need to be given rewards at all? If they are genuine loyalists, they should turn out of their own accord and vote for their own party or candidate without being enticed with a benefit (Stokes et al., 2013: 111). This widely held assumption, however, isn't supported by the facts in the Indonesian case. As alluded to earlier, detailed studies of grassroots electioneering for the 2014 elections by Aspinall and his colleagues (2017: 2), for example, capture candidates' "deep anxiety about the reliability of even supposedly loyal voters, suggesting that their votes were vulnerable if not secured with a payment." Hence, the loyalty of the so-called loyal voters cannot be taken for granted since their allegiance to the candidate tends to be affected by short-term electoral incentives. In short, candidates face two related but distinct risks from targeting swing voters. First, that swing voters might take the money but not vote for them. Second, that a core voter who doesn't get funds will defect. The latter seems the bigger risk for candidates, which calls into question whether it is appropriate to call a group of voters that candidates target as a 'core.'

It is worth noting that my finding in Indonesia differs from those in the study of Diaz-Cayeros and his colleagues (2012) in the case of Mexico. They found that political parties favoured their partisan voters as an artefact of endogenous party loyalty. Diaz-Cayeros et al. (2012: 23) argue that "if swing voters are constantly targeted with benefits, core voters will no longer tolerate it, and will soon become open to mobilisation by other political parties, behaving much like swing voters in future elections." In their theoretical framework, party loyalty is conditional, or endogenous, rather than fixed. The Indonesian case is different. The driving factor behind candidates' insistence on giving cash to their 'loyal' supporters is suspicion of the allegiance of their perceived loyalists in the context of increasing electoral uncertainty induced by intense intraparty competition. In addition, as noted above, voter loyalty in Indonesia isn't framed in terms of partisan convictions, as the literature (including Diaz-Cayeros et al., 2012) conventionally assumes. Given that beneficiaries are in practice largely selected in terms of personal network relationships, we witnessed why the targeting of so much vote buying ends up in the hands of undecided voters who don't always reciprocate with votes.

#### 5.5 Conclusion

So, how do politicians in Indonesia determine which voters to target? When it comes to the targeting strategies, the existing literature offers two conflicting schools. As we have seen, one school of thought holds that in terms of distributing cash payments, political machines favour their core supporters over ideologically indifferent voters (e.g. Nichter, 2008; Diaz-Cayeros et al., 2012; Stokes et al., 2013). Another school claims that parties will not squander their limited budgets on core supporters, but instead expend them on swing voters in an attempt to convince them to support the giver (e.g. Lindbeck and Weibull, 1987; Dixit and Londregan, 1996; Stokes, 2005).

At first, my findings seem to show that candidates and intermediaries tend to target constituents who they think of as being truly loyal, strongly evocative of core-voter strategy. Moreover, the logic of various elements of Indonesia's institutional framework, such as optional voting, open-list PR, and ballot secrecy, provides strong incentives for political machines to favour party loyalists because such voters are thought of as being more reciprocal, and as a more predictable source of votes. Yet, theoretically, the model is built on the assumption that voters can be categorised 'loyalist' as long as they are proximate to a party in ideological or partisan terms (Stokes et al., 2013: 45). However, the number of partisan voters is limited in Indonesia, and this small segment of voters is highly contested among co-partisans in the context of open-list PR which incentivises zerosum, intraparty competition.

If candidates and broker really favour only their loyal party supporters, as sometimes seems suggested by how they describe their strategies, how could they expect to win, given the limited number of such voters? In an attempt to address this puzzle, this chapter offers an alternative explanation to the literature by highlighting the importance of personal networks in explaining clientelist strategies in Indonesia. My argument works for a context where partisan ties are relatively weak, where the electoral system is candidate-centric, where the organisation of vote buying isn't partybased, and where personalised loyalties matter far more. Given the party's captives are both limited and contested among internal rivals, candidates skilfully use their personal networks as the major tool of voter mobilisation. These personal networks have multiple dimensions ranging from kinship, ethnic, and religious ties, to patronage to brokerage networks. In short, the empirical evidence paints a picture largely consistent with the personal loyalist (i.e. people who were selected through personal networks) strategy I highlight.

The reliance on such networks, however, makes the personal loyalist strategy vulnerable to the problems of targeting and principal-agent breakdown that arise in the relations between candidates, brokers, and voters. In particular, this chapter has identified a pattern in the allocation of vote buying where, despite politicians' and brokers' strongly expressed preference for targeting loyalists, they end up distributing to voters who are in fact not loyal to the candidates. I have discussed three major difficulties associated with the personal loyalist strategy: first, candidates and brokers misidentify the number of loyal voters because they tend to mix up partisan and personal loyalties. Second, the loyalty concept in Indonesia is ambiguous and has multiple dimensions relying on a more personalised rather than strictly party-based relationships. Unsurprisingly, when it comes to determining who gets targeted for vote buying, candidates depend on personal networks rather than judging their targets in terms of partisan and ideological leanings. Third, agency loss also explains why the targeting of vote buying ends up with voters who don't reciprocate. Candidates obviously know that brokers have an interest in exaggerating their influence. They are also aware that voters whom brokers classify as loyal are often typically contested by brokers working for rival candidates and often receive multiple payments. Even so, precisely because political actors aren't really confident about the 'loyalty' of their so-called loyal voters, they typically fear many such voters will not vote for them unless they receive gifts.

In conclusion, the targeting strategy in Indonesia must be seen as a distinctive type that functions like neither the core-voter nor the swingvoter model. Much of the literature on core- and swing-voter models is framed by a context different from that in Indonesia. The existing literature relies on the underlying assumption that party machines are both engaged with voters and capable of mobilising them. This sort of analysis doesn't fit the Indonesian context, where elections have been largely driven by candidate-centred politics in which a lot of the connections with voters aren't mediated by parties, but instead by informal brokerage networks. The differences in context make the dynamics of vote buying, at least the targeting, very distinctive.

#### Bibliography

Books, Journal Articles, and Unpublished Papers


**Open Access** This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

# Vote Brokerage, Personal Networks, and Agency Loss

In the previous chapter, I showed that the targeting strategy used by Indonesian legislative candidates that combines the party loyalist strategy and personal networks makes vote buying tremendously inefficient. Candidates and brokers plan to target party loyalists, but in reality, most vote buying happens among uncommitted voters. In this chapter, I argue that such vote-buying efforts not only suffer from inefficiency of delivery, but are also undermined by a substantial amount of 'leakage,' given the intrinsic principal-agent problems that arise in the relations between candidates and brokers. I show that most candidates lack mechanisms to systematically monitor or discipline their brokers, making vote-buying attempts vulnerable to broker predation. I also provide empirical evidence of rent-seeking behaviours by brokers, drawing on a unique survey of brokers.

But if vote brokerage is subject to rent-extraction, this doesn't mean that all brokers are equally untrustworthy and unreliable. The available body of research on electoral clientelism in Indonesia clearly demonstrates that vote brokerage—which relies on personal networks—remains critical to electoral success, given brokers' significance in expanding candidates' electoral base and distributing material inducements. Overall, I argue that despite the inherent weaknesses of vote buying using personal networks, these networks really matter in shaping how vote buying works, which brokers are recruited, and how targeting occurs. This chapter accordingly aims to explore brokerage networks—who the brokers are and what drives them to join a candidate's success team. It also discusses the logistics of vote buying, such as how candidates recruit and monitor brokers and how brokers determine what, when, and how much money to give. It seeks to answer these and related questions through an examination of my field research findings and data produced by my broker and candidate surveys, as well as voter surveys from Indonesia's 2014 elections.

# 6.1 How Many Brokers Are There in Indonesia?

My unique survey of a probability sample of vote brokers provides information about the estimated population of brokers in West Sumatra, East Java, Central Java, and North Sulawesi. As I explain in Chap. 1, the selection of these four provinces was primarily guided by the fit of cases with the statistical findings of large-scale surveys in 73 out of 77 electoral districts across Indonesia. I then selected these four provinces to reflect variations in levels of mass partisanship and vote buying across Indonesia.

Table 6.1 illustrates that relative to other provinces, Central Java is slightly wealthier and has slightly better life expectancy and human development scores than the average, but in terms of literacy rates and number of school years completed, it is below the national average. In East Java, the patterns are broadly similar to Central Java, with the exception in terms of life expectancy; East Java is marginally behind the national average. Although much of North Sulawesi and West Sumatra do relatively


**Table 6.1** Demographic profile of four provinces

Source: http://data.go.id/dataset/ipm-dan-komponennya-per-kabupaten

Note: All data from 2012

well in terms of life expectancy, literacy, number of school years completed, and human development scores, both score badly on average monthly per capita expenditure. In socio-cultural and political terms, Central Java is widely known as the secular PDI-P's primary stronghold with a mixed population of *abangan*, socio-economically lower-class nominal Muslims, as well as more pious *santri* Muslims with traditionalist Islamic backgrounds (Geertz, 1960). While East Java is a centre of NU-style traditional Islam, it also has a large concentration of *abangan* in the southern part of the province. West Sumatra is the home of Islamic modernists, who adhere to orthodox Islam while accepting modern ideas, and it provides the strongest support outside Java to the biggest modernist organisation, Muhammadiyah. North Sulawesi is predominantly Protestant with a sizeable Catholic minority and is recognised as a stronghold of both PDI-P and Golkar. In terms of political alignments, by contrast, East Java and West Sumatra are categorised as 'swing' regions with no single party dominating.

As already explained in Chap. 1, my research began with a face-to-face survey of a probability sample of 299 elected candidates for provincial and regent/district DPRD electoral districts in four provinces. Then, for every randomly selected DPRD member, three of their brokers who helped them during the 2014 election were also randomly selected, making about 900 brokers in total interviewed. The population of brokers in each region was, of course, unknown because brokerage teams are informal and unregistered. To determine the broker population, I made an estimate based on the average number of brokers mentioned by local MP respondents. Hence, the sample in four provinces reached a total of 1199 respondents consisting of 100 provincial DPRD members, 300 brokers of provincial DPRD candidates, 199 Regent/City DPRD candidates, and 600 brokers of Regent/ City DPRD candidates (one sampled district DPRD member cannot be interviewed within the field duration, but his brokers were given permission to participate in this survey). I personally administered these unique, representative surveys of elected candidates and brokers from September to October 2014 (for more information about these surveys, see Appendix B). To my knowledge, mine is the largest and most detailed survey of candidates and brokers undertaken to date anywhere in clientelism research.1

<sup>1</sup>For comparison, Stokes and her collaborators conducted a survey in 2009 of about 800 councillors and non-elected brokers in four Argentine regions—the provinces of Cordoba, Misiones, San Luis, and the Greater Buenos Aires (see Stokes et al., 2013).

Table 6.2 presents the survey results, showing the average number of brokers by province. There is considerable variation. In each province, the average number of brokers for provincial legislature (DPRD I) candidates is greater than the average number for district legislature (DPRD II) candidates. This gap is largely because the electoral districts at the provincial level are larger and their populations are greater, which in turn necessitates a larger number of brokers. It is important to note that in Java, the boundaries of the national and provincial electoral districts coincide (Aspinall and Sukmajati, 2016: 15). Broker armies in both Central Java and East Java were large, and vote-buying strategies were extensively pursued in both provinces. This finding contrasts with the normal expectation. Previous works argued that the greater the size of the electoral constituency, the lower the likelihood of clientelist strategies being pursued (Stokes, 2007; Hicken, 2007). However, my survey findings pointing to large brokerage networks in Central and East Java correspond with general findings from my pre-election surveys in all electoral districts of these two provinces, which suggests that voters in these provinces were more likely to view vote buying as a normal practice (see Chap. 1).

Consistent with prior expectations, the brokerage networks in West Sumatra were much simpler relative to other regions. Instead of utilising large networks of brokers for grassroots campaigning, most elected candidates in the province instead drew on traditional leadership structures, known locally as *tungku tigo sajarangan* (the three stoves at the hearth), that comprises three pillars: *adat* or clan elders (*ninik-mamak*), religious leaders (*alim-ulama*), and *cadiak pandai* (enlightened intellectuals). Where they could, candidates approached and recruited these people into their personal teams (Edi Inrizal, Interview, 21 September 2014). This qualitative evidence parallels my voter survey results in West Sumatra, indicating that the level of vote buying in the province was much lower than in other regions in Indonesia, presumably largely because candidates were instead relying partly on the informal influence exercised by these community leaders.


**Table 6.2** The average number of brokers per candidate


**Table 6.3** Estimated broker population in four provinces

To estimate the broker population, the average number of brokers per legislator is multiplied by the total number of legislative seats in the province concerned. For example, Table 6.3 shows that the average number of brokers for each provincial legislator in West Sumatra was 44.3 people. I then multiplied these with the total seats so the estimated broker population in the province was 2880. In four provinces, it is estimated that there were around 97,657 brokers working for successful provincial legislative candidates in the 2014 elections, and 576,922 brokers for successful candidates in district legislatures. Note that this is only a calculation that includes the successful candidates. Brokers who worked for candidates who didn't get elected aren't included. Hence, in fact, the total number of brokers must have been much higher.

For the sample to be representative, I then weighted the data using the following formula:

$$\mathbf{W}\_{\mathfrak{g}\mathfrak{a}} = \frac{N\_{\mathfrak{z}\mathfrak{k}}}{N} / \frac{n\_{\mathfrak{z}\mathfrak{k}}}{n}$$

in which:

*Wijk = Weighting variable for provincial data i, DPRD level j, broker category k*

*Nijk = Population amount for province i, DPRD level j, broker category k nijk = Sample amount for province i, DPRD level j, broker category k N = Total population n = Total sample i = West Sumatra, Central Java, East Java, North Sulawesi j = DPRD I, DPRD II*

*k = Legislators, brokers*

To determine statistical significance, a standard error value is needed. In this study, the standard error is calculated by assuming stratified random sampling: population grouped by province, office level (i.e. DPRD I and DPRD II), and respondent category (i.e. candidates and brokers), and then the sample in each stratum is selected by simple random sampling. However, the standard error of these surveys could have been larger than estimated. This is largely because, in practice, the sampling is conducted with a more complex procedure. All these sampling and statistical procedures are described further in Appendix B. Based on the available data, it can be concluded that the total number of brokers working for successful provincial and district legislative candidates in four provinces in 2014 was quite large (674,579 people). In the 2014 legislative election, there were 62,994,652 registered voters in these provinces. Hence, a broker-to-voting-eligible-population ratio in the four provinces was around 1 broker to 93 voters. Of course, if we include those who worked in other provinces, including those who helped unsuccessful candidates, the estimate would have been much bigger.

### 6.2 Who Are the Brokers?

My survey of a probability sample of brokers offers insight into the characteristics of brokers. In terms of gender composition, Table 6.4 clearly shows that political brokerage is a male-dominated field. Most of the brokers interviewed in four provinces are male, with higher percentages in Central Java and East Java at 92.7% and 91.4%, respectively. The majority of the surveyed brokers were aged between 41 years and 55 years. The average age of brokers was 44 years (the youngest broker was 17 and the oldest 76). Candidates presumably tended to recruit middle-aged men because the middle-age period is associated with the greatest independence, esteem, prestige, and social involvement; therefore, these brokers can demand respect from both the young and the elderly in the community (Martel, 1968: 56)—a necessary condition of successful brokerage activity. The majority of the interviewed brokers (77.6%) were Javanese due to the fact that one-half of the survey interviews were conducted in East Java and Central Java. Brokers were mostly Muslim (91%), with only 9% of the respondents being Christian.

How do the findings place brokers with respect to the general population? To provide an apple-to-apple comparison, I use voter survey data only from East Java, Central Java, West Sumatra, and North Sulawesi. In



(*continued*)

178B. MUHTADI


**Table 6.4**

(continued)

conducted during 2–9 June 2014; and (3) by SMRC during 20–24 July 2014. The number of samples in the Indikator June survey, the SMRC June, and SMRC July surveys analysed from four provinces (West Sumatra, East Java, Central Java, and North Sulawesi) used in this study was 420, 780, and 400 respondents, respectively. Meanwhile, candidate and broker data are taken from my survey of low-level politicians and brokers, 30 September–25 October 2014 order to increase the sample size, I utilise three nationally representative voter surveys that were held simultaneously just before the broker survey was conducted (September–October 2014). In terms of socio-economic status (SES factors), my survey results reveal that, compared with voters in four provinces, brokers tended to be more educated and slightly wealthier. Table 6.4 also shows that most voters were employed in blue-collar jobs (82%) such as farmers, fishermen, small street-stall sellers, and so on, compared with 63.8% of the sampled brokers whose jobs can be grouped as blue-collar employment. In terms of monthly earnings, surveyed brokers weren't as a group poor by Indonesian standards, though they were also not wealthy. The modal respondent earned between IDR 1 million and less than IDR 2 million per month (38.4%), though 36.6% of brokers said their monthly income was more than IDR 2 million per month. In contrast, almost half of the voters reported monthly earnings of below IDR 1 million. Brokers were also more educated than the general population. The modal respondent in my brokers' survey was a senior high school graduate (45.8%), though 19.9% of brokers said they were college graduates. Almost half of the sampled voters had a maximum of primary education or no education at all.

When asked whether politics was the brokers' main source of income, almost all respondents (99%) said that politics wasn't their main profession. The largest main occupation mentioned was farmer/animal breeder/ fisherman (26%), followed by private employee/self-employed and bluecollar worker/maid with 10% and 9%, respectively. A significant number of brokers described themselves as entrepreneurs/businessmen (8%), streetstall sellers (7%), and teachers/lecturers (5%). Some brokers derived income from automotive services (4%) and wholesale trade (3%), were retirees (3%), or village bureaucrats (2%). The remaining brokers were professionals (lawyers/doctors/etc.), drivers, security personnel, freelancers, or civil servants (*pegawai negeri sipil*, PNS).

Clearly, most of the surveyed brokers weren't full-time professional brokers; instead, brokerage was simply a temporary or part-time job at election time. This finding contrasts sharply with Zarazaga's description (2014) of everyday political mediation in Argentina, which indicates that brokers perform a wide range of roles to help their party, including performing constituency service and performing basic governmental tasks. In Indonesia, brokers are generally not multitasked in this way, but instead focus on helping their bosses to win an election. Their challenge is to combine this job with their everyday occupation.


**Table 6.5** Logistic regression of whether or not the respondent was a broker

Note: \**p* < 0.05; \*\**p* < 0.01; \*\*\**p* < 0.001

In terms of social involvement, the mean of the brokers' responses to questions asking about their involvement in various forms of associational life was 17.94 with a standard deviation of 3.68. Overall, brokers were highly socially engaged people compared with the general population. For instance, 50% of the brokers described themselves as being active members of local religious community groups (*majelis taklim*), which was the largest civic association mentioned, compared with 23% of the voters reported being active members of such group.2 Note that each respondent was allowed to provide multiple answers. The next most important association was rotating credit groups (*arisan*), with 34% of the brokers reported being active members, compared with 25% of respondents who said so in my voters' survey. Brokers were also more actively engaged than voters in some organisations such as NU, youth community councils, agricultural groups, or industrial unions and cooperatives.

To further compare the characteristics of brokers with the general population, I conducted a case-control study by combining data from my unique dataset of brokers and voter surveys (Table 6.5). In doing so, people with the outcome of interest (i.e. brokers) are compared and matched

<sup>2</sup>This is based on my post-election survey of voters in 2014.

with a control group (i.e. general voters).3 The assumption here is that our respondents in the voter opinion surveys weren't brokers. The results suggest that, compared with voters based on data from my polling institute's voter surveys, brokers were older, much more likely to be male, had more education, and enjoyed higher median household incomes than the typical Indonesian, and they tended to be more active in social organisations. In determining the number of selected matching variables, the region variable (North Sulawesi and otherwise) is set as a control. So this is not possible determinant of becoming a broker, given we have selected the variable from the beginning. We included the variable of region in the equation to control the variable of religion because North Sulawesi is feared to be biased for non-Muslims, while West Sumatra, East Java, and Central Java are potentially biased for Islam.

### 6.3 How Do Candidates Recruit Brokers?

Most candidates saw broker recruitment as critical to their strategies. As one candidate of the Islamist party PPP, who ran for re-election in 2014, explained: "The last elections were all about a distribution strategy. Therefore, how we selected success team members was a key to success in securing victory" (Interview, 21 April 2014).

But what sorts of persons did candidates recruit? Generally, individuals recruited into campaign teams could be divided into two groups: party functionaries (either at the sub-district or at the village/precinct level) and non-party members (Kadir Karding, interview, 18 April 2014). Candidates felt they must involve local party functionaries in a success team, not only for practical reasons but also to avoid offending their fellow party members. Since party structures were contested by competing candidates on the party list, however, candidates couldn't rely exclusively on party cadres. Some candidates accused party functionaries of operating as 'double agents' (*main dua kaki*) or 'money grubbers' (*mata duitan*).4

In terms of non-party brokers, my broker survey provides a complete picture of how the process of recruitment works. Figure 6.1 reveals that brokerage structures take advantage of relational networks (Aspinall et al., 2015). Asked about their relationship with the person who asked them to join the campaign team, 79% of brokers were recruited by those they cat-

<sup>3</sup>For further discussion on case-control study, see Alan Agresti (2007).

<sup>4</sup>My informants spoke on the condition of anonymity due to the sensitivity of this information.

**Fig. 6.1** Brokers' relationship with the person who invited them to join the team (%). Source: My survey of brokers, 30 September–25 October 2014

egorised as family and friends. Respondents were allowed to give multiple responses, but this figure clearly indicates that candidates rarely recruited people they didn't already know well as core team members, who in turn recruited close associates, and so on down the campaign team structure. The most frequent explanation candidates gave for prioritising relational networks was that brokers with whom they were personally close were more loyal and less likely to shirk (Aspinall et al., 2015). Also, recruiting through personal networks helped to reduce the costs of maintaining a network of brokers because members would be willing to work hard on the campaign even without payment, or for lower payments than otherwise. In contrast, a modest portion of brokers were recruited by their party leaders. Interestingly, when it came to broker recruitment, candidates were more likely to favour neighbours over organisational networks, confirming that brokers were genuinely recruited on the basis of personal connections.

Consistent with earlier works, we also found evidence that most candidate-broker relationships in Indonesia were highly personalised. About 61% of brokers were asked to join by the candidate directly. This figure more than tripled the findings of a survey conducted in in Central Java III constituency by Aspinall and his collaborators (2015). Note, however, that the population of my survey was brokers who worked for candidates who were *elected* at provincial and regent/mayoral levels. It may be that candidates who constructed success teams on the basis of intimate relationships were more likely to win their seats.

Meanwhile, only 38% of respondents were asked to join the success team by other, more senior brokers. These respondents were asked to specify the position in the campaign team structure of those who had approached them to join. The modal answer to this question was "village coordinator" (34%), followed by "sub-district coordinator" (30%) and "base-level brokers" (22%). The remainder identified "regency/city coordinator" or "provincial coordinator." This parallels prior research that when candidates exhaust their supply of close friends and family members when forming a success team, they typically turn to campaign coordinators to recruit the rest. Unavoidably, among those recruited in this way, a considerable portion have connections that are more transactional than personal (Aspinall, 2014; Aspinall et al., 2015).

In short, the pattern of relationship between candidates and brokers reflects a centrality of personal networks and, by extension, relatively marginal role of political parties, in grassroots campaigns in legislative elections. Asked whether they knew anything in advance about the candidate they supported, 71% of brokers knew the candidate personally and were close to him/her. Around 19% knew the candidate personally but weren't particularly close before becoming part of the campaign team. Only 4% had heard about the candidate but never met him/her; 5% had never heard of the candidate before. Again, recall that the sampled brokers were those working for *elected* candidates. Prior contacts between candidates and brokers partly mediate electoral success. Many candidates failed to be elected presumably because they used a much looser method of selecting brokers without ensuring whether or not they were truly committed to them.

Such prior contacts, however, don't necessarily imply long-term clientelist interactions. Some brokers might have personal affective ties to the candidate, without clientelist exchanges. Respondents were asked to name whether they had had prior assistance from a number of actors such as the candidate, success team village coordinators, sub-district coordinators, district coordinators, and party leaders. Although 71% of brokers knew the candidate personally prior to becoming a member of his/her success teams, the left panel of Fig. 6.2 shows that only 37% had previously received assistance from the candidate or team members. As a follow-up, these 37% of brokers were requested to specify the forms of assistance they

**Fig. 6.2** Prior mutually beneficial exchanges involving brokers (%). The question in the left panel reads: "Before becoming a success team member, have you ever received assistance (e.g. in government paperwork, job assistance, project access, schools, health care treatment for you or family members), from the following parties?" If "Yes," the interviewer asked a follow-up question: "In what forms? (Can be more than one answer)." The responses to this question are then shown in the right panel. Source: My survey of brokers, 30 September–25 October 2014

had received. As shown in the right panel of Fig. 6.2, the modal answer was 'project access' (66%), such as constituent funds for the construction of roads, bridges, and so on.

Clearly, distribution of such patronage helps create networks for the candidate (see Chap. 5). It gives incumbent candidates a comparative advantage in getting re-elected since they have had the opportunity to form patronage-based networks since at least the preceding election and have greater access to pork-barrel projects and other state resources while sitting in the legislature. Numerous brokers had also received educational help (15%), job offers (12%), health-related assistance (9%), and help in government paperwork (9%). The proportion of those who said they had received other forms of help, such as money, goods, or protection, was also numerically significant (19%). Meanwhile, 63% of brokers hadn't received help from the candidate. These respondents mostly were located at the ground level and didn't work directly with the candidate. They interacted directly with voters instead. I call them 'extended success teams'—people who are recruited by the senior brokers who usually enjoy direct access to the candidate. As explored in the following section, these extended success teams are a source of rent-seeking because brokers at this level often have transactional rather than personal commitments to the candidate.

However, an important point is that the left panel of Fig. 6.2 demonstrates that brokers seem to be more tightly bound to candidates than with other actors when it comes to clientelist exchanges. Only a tiny fraction of brokers had had prior clientelist interactions with more senior brokers. Of those, only an insignificant number had benefited from party leaders, confirming that the relationship between candidates and brokers is more personal than partisan. The results show evidence of the weakness of clientelist ties between party structures and brokers. Likewise, very few respondents mentioned their prior exchanges with campaign coordinators at the district level (1%), sub-district level (1%), and village level (4%). All of this suggests that the nature of the relationship between ground-level brokers and coordinators at district or sub-district level is generally one-off, noniterative, and short term. The relationships between candidates and at least senior brokers, however, tend to be more personal and long lasting.

#### 6.4 What Are Brokers Looking for in a Candidate?

The dominant literature on clientelism suggests that an individual's motivation to join a broker network is primarily driven by partisan orientations (e.g. Auyero, 2001; Zarazaga, 2014). The scholarly focus on the role of political parties and party brokers has left the function of individual candidates and their non-party brokerage networks relatively understudied. But in a context like Indonesia, where parties play a minor role in grassroots campaigning, and where under the open-list PR system co-partisans compete against each other, why should a broker end up working with one candidate rather than another? One answer is that brokers vary by types depending on the differing commitment and the degree of personal closeness to the candidate. Some are "pragmatists," in Szwarcberg's (2015: 2) term, or "opportunists" in Aspinall's (2014: 545) categorisation who seek immediate payoffs during the course of a campaign, and will thus gravitate towards whichever candidate offers the best material payoffs. Others are "clientelist brokers," in Aspinall's term, who intend to have durable relations with the candidate in the hope of receiving future benefits (2014: 545).

My own findings suggest, first and foremost, that a candidate's personal reputation is important in attracting brokers to a success team. Figure 6.3

**Fig. 6.3** Reasons for joining a candidate's success team (%). Source: My survey of brokers, 30 September–25 October 2014

illustrates the main reason respondents gave for joining a success team with the important caveat that the sampled brokers here all worked with successful candidates. Few depicted themselves as pragmatist brokers whose materialistic motivations are typically blamed for the failures of success teams to reach expected vote targets. Only an insignificant number were primarily motivated by ethnic and religious factors. The primary motivations success team members gave were the 'candidate's *ketokohan*' (personality or personal qualities) and the 'candidate's programmes.' These responses point to the significance of personal reputation in generating electoral support independent of the party. This finding corresponds with the dominant discourse among candidates and brokers in interviews during my 13-month fieldwork in Indonesia. They sometimes used the word *figur*, which is more or less synonymous with *ketokohan*, to illustrate that it was the candidate's individual qualities that really mattered under the open-list PR system.

Virtually all brokers admitted that a candidate's personal reputation helped increase his or her vote share. My very large dataset surveys of voters in 73 electoral districts across Indonesia also reveal that, when asked to name the personal qualities they think are most important in a candidate, the majority of respondents named "honest/trustworthy" and "cares for the people." In Indonesian politics, caring for the people connotes having a track record of having delivered benefits or patronage to voters. Similarly, the term 'programme' as in Fig. 6.3, has a specific meaning in the Indonesian context, as Aspinall et al. (2015: 12) explain:

It refers not to a collection of policies or ideas the candidate supports—what we might elsewhere call a candidate's platform—but to what a candidate has done for the village/voters/constituency. In other words, "program" refers to how adept or generous the candidate has been in providing the area with pork, patronage, and club goods.

Candidates who are able to meet those desirable criteria come to be viewed as having higher prospects of victory. Other things being equal, brokers prefer "to work for a candidate who was likely to win rather than one who was personally wealthy but had poorer prospects of victory" (Aspinall et al., 2015: 12). This statement nicely parallels my interviews with brokers: when they decided to work for a candidate, it had little to do with whether the candidate was using a vote-buying strategy, and much more to do with the candidate's reputation, since a strong reputation would maximise chances of victory. This isn't to say, of course, that personal reputation is itself sufficient to obviate the need to distribute cash to voters; even many candidates with strong records of patronage delivery and personal popularity in their electoral districts felt that they still needed to use vote-buying strategies in order to lock in the support of their voters.

# 6.5 Whom Do Brokers Target and How?

What structures do politicians construct to engage in vote buying? The typical organisations of brokers, which are known locally as *tim sukses* (success teams), work within a pyramidal and territorial structure. At the apex of a typical success team for a DPR member is an inner circle of two or three people who help build and control the network. As observed by Aspinall (2014), the bulk of the structure consists of a territorially organised network, stretching down to the village and, often, hamlet or even polling booth level, where local brokers are recruited to directly influence voters (Aspinall and Sukmajati, 2016). Hence, the majority of success team members consist of base-level brokers—what I called the extended success team—whose local networks allow them to collect votes on the candidate's behalf. These base-level brokers play a critical role since they have to provide lists of voters willing to vote for the candidate, deliver the payments to them, and then monitor the recipients to ensure they show up on voting day.

In terms of targeting strategies, it is clear that when it comes to determining who gets targeted for vote buying, brokers predominantly rely on personal networks or simply on personal closeness to brokerage networks. As shown in Fig. 6.4, brokers usually target people among their family members, neighbours, friends, and relatives in the village. Although brokers have an incentive to expand electoral support beyond their personal networks, they tend to favour easy targets to increase the effectiveness of vote buying and reduce the cost. When asked to name multiple targets, a modest portion of brokers also mentioned those who were active in their professional and religious community groups. When asked to specify where their vote targets lived, most of them said they resided in the same neighbourhood or village (Fig. 6.5). There was little variation between provinces, indicating that the targeting strategy based on brokers' personal networks is likely common across Indonesia.

**Fig. 6.4** Who did brokers ask to vote for the candidate they support? (%). Source: My survey of brokers, 30 September–25 October 2014

**Fig. 6.5** Where do the people targeted by brokers live? (%). Source: My survey of brokers, 30 September–25 October 2014

A parallel discourse emerged from my qualitative findings, suggesting that most brokers chose to prioritise voters whom they know and who know them. However, the number of such voters was limited. Recall that when brokers often talked about family, they didn't only refer to their immediate kin but much wider ties, linked to them via kinship or common ancestors, marriage connections, emotional relationships, and place of birth or residence (Alamsyah, 2016: 108). In my interviews with many brokers, I asked why they favoured such a personal strategy. The most frequent answer was that such persons were the most reliable voters. By directing benefits to their personal networks, they believed their offers wouldn't go to waste. In fact, as long as the personal connections between brokers and voters were strong, the payment was simply often just a courtesy. There is a shame effect if the targeted family members and close associates accept the offer but don't reciprocate with votes. One broker working for a candidate running in an electoral district in East Java said, "The point isn't *how much* money did you give but *who* gave the money to voters. Voters in the villages don't care about politics," (Interview, 30 September 2014). Another broker, whose wife was running for a district parliament seat in South Tangerang, put it, "They even don't know which candidate I support" (Syafrani, Interview, 5 December 2014). The broker wasn't only an influential leader in the district but also one of the leading commentators on law and politics, appearing on prime-time television programmes several times a week.

All of this suggests that it isn't the message but the *messenger* that matters. As one broker put it, "The more brokers a candidate has, and the greater family and friendship networks a broker has, the bigger chance for a candidate to win" (Interview, 4 October 2014). My low-level politician survey found that the number of brokers working for the sampled candidates, who were successfully (re)-elected in 2014, varied in size, including one extreme outlier—one candidate reported working with 3000 brokers. But the mean number of brokers was 149.44 with a standard deviation of 299.1, suggesting that team sizes were quite large. Faizin, a successful candidate from the Islamist party PPP who ran in 2014 for a district parliamentary seat in Batang, Central Java, explains:

For the candidates, rather than assigning brokers to collect a significant number of votes … beyond their limits, it is better to have a bigger army of brokers but ask every one of them to chase fairly modest vote targets. If you burden your brokers with impractical tasks, they will certainly fall short of expectations. (Interview, 23 January 2017)

In other words, if the target is set too high, it will be difficult for brokers to reach it due to the limits in the number of family-and-friend votes they can access. But since most candidates need to chase a large number of votes, brokers in fact often do exhaust the support of those with close personal connections to them. Inevitably, they must reach out to those with whom they lack close personal ties, increasing the risk of wastage and slippage. The longer the list a broker collected, the more likely it would include voters beyond his/her immediate circle, which then increases the prospect of a higher failure rate.

# 6.6 How Do Brokers Determine What, When, and How Much to Give?

It is technically difficult to distribute benefits to a large number of voters when time is running short, as it inevitably does towards the end of a campaign period. Part of the solution is that candidates favoured distributing cash over staple foods because it is extremely difficult to procure and distribute food in a context where vote buying is illegal. The results from the brokers' survey show that they were three times more likely to report handing out cash (16%) than foodstuffs (5%), which is consistent with our voter survey results. Designed as a multiple choice question, the most common item offered to those who self-reported vote-buying attempts (25.1%) was money (75.5% of 25.1% = 18.7%); foodstuffs (12.8%) such as rice, sugar, or noodles; and household items such as kitchenware or crockery, or religious clothes like headscarves, prayer robes, or mats (11.4%). Some respondents also mentioned other items such as clothing, cigarettes, health and death insurance, medicine, and so on.

This study did find substantial heterogeneity by province where the proportion of brokers reporting engagement in distributing cash was 22% in Central Java, 14% in East Java, 8% in North Sulawesi, and only 4% in West Sumatra. This closely mirrors the data from my voters' surveys suggesting that vote buying has become an endemic problem in Central and East Java but is much less common, especially, in West Sumatra. We must also assume a significant social desirability bias, leading to some underreporting of engagement in vote buying, in each province.

The literature on clientelism has long stated the effect of constituency size on the relative costs and effectiveness of vote buying. Scholars argue that a small constituency size may increase the likelihood of vote buying (Stokes, 2007: 86–87; Hicken, 2007: 56–57). However, my broker survey shows that constituency size doesn't seem to have much effect5; in fact, 24% of provincial brokers reported they engaged in vote buying, while only 20% of district brokers said so, though this difference wasn't statistically significant. At least this suggests that vote-buying efforts didn't only take place in small electorates but also happened in larger electorates. When it comes to the amounts of money distributed, there was significant variation, including inter-regional variations. For instance, candidates spent lesser sums per voter in Java than in the two outer island provinces. In many instances, it is more expensive to 'buy' urban dwellers than rural residents. To some extent, there are also inter-candidate variations. In some areas, wealthy candidates set the price of a vote much higher than the market price (*harga pasaran*) of a vote (Aspinall et al., 2017: 5).

Overall, however, when it comes to the amount of cash distributed to voters, the effect of constituency size really mattered. Figure 6.6 illustrates the pattern. Brokers working with provincial candidates usually distributed smaller sums of money (typically below IDR 25,000 per voter) than district-level brokers. It shows those who gave out larger sums, from IDR 30,000 to IDR 50,000 to be exact, were more likely to be district-

5This finding isn't necessarily inconsistent with the constituency size argument, however. It isn't the number of voters that matters, but rather, votes/seats, or the number of votes needed to win. If the number of seats grows as the number of voters grows, then there is no change in the size of the constituency.

**Fig. 6.6** How much money did brokers distribute to voters? (%). Source: My survey of brokers, 30 September–25 October 2014

candidate brokers than provincial brokers. Through interviews with candidates running for the DPR, I ascertained that they mostly gave out even smaller payments than did candidates for the provincial DPRD I seats. Albeit with significant variations in the amounts paid, they generally distributed smaller cash handouts that ranged between IDR 10,000 and IDR 15,000 per voter. The reason is simple: candidates running for national-level seats had to collect many more votes than did candidates for the provincial-level parliaments.6 Candidates contesting for national DPR seats had to win hundreds of thousands of votes to secure victory, while those running for provincial DPRDs usually had to secure tens of thousands of votes. Similarly, candidates for district-level legislature had to provide much larger payments because they had to secure only 2000–10,000 votes.

In sum, the total expenditure on vote buying of national and provincial candidates was still more costly than that of district candidates. A successful national candidate from the traditionalist Islamic party running from one constituency in Central Java admitted distributing 450,000 envelopes containing sums of between IDR 10,000 and IDR 20,000 each (Interview,

<sup>6</sup>However, this is not the case in Java where the boundaries of the national and provincial electoral districts coincide (Aspinall and Sukmajati, 2016: 15).

20 April 2014). If the average price of the vote was IDR 15,000, the candidate spent IDR 6.75 billion on vote buying. A re-elected candidate, from PAN, admitted that running in the 2014 elections was extremely expensive. He said, "I ran three times in the elections, but the 2014 campaign was the most 'brutal' one" (Interview, 22 April 2014). He was visibly emotional and appeared to tear up when recalling his brother's crying during his campaign because he felt guilty for his inability to help support him financially, instead assisting him only through his prayers.

Given there are overlapping constituencies in Indonesian legislative elections, politicians reduce the costs associated with the larger electorates by coordinating their efforts through cooperative deals, which are usually called 'tandem arrangements' with other candidates. My unique survey of candidates shows that provincial candidates were more likely to run in a tandem pair with a DPR candidate than with district candidates. About 55% of the sampled district candidates reported collaborating in a tandem with provincial candidates, while 73% of the surveyed provincial candidates ran in a tandem pair with district candidates. The data show that provincial candidates were more likely to have such cooperation than district candidates. Almost all of these tandem arrangements occurred within a single party. In terms of vote-buying expenses, DPR candidates usually shared funding with lower-level candidates who were more responsible for providing personnel to conduct the transactions on behalf of either lowerlevel themselves or the DPR candidate. In my interviews with DPR candidates, they were concerned about the reliability of their local counterparts in handing out cash on their behalf. It was often the case that tandem arrangements benefitted only the local candidates, who used the extra money that came to them from the DPR candidate to increase the amounts distributed to each voter without acknowledging their higher-level counterpart.

Figure 6.7 provides insights into the timing of cash distribution. Using a disguised technique to reduce social desirability bias, I asked the respondent to observe how many brokers from the same village as him/her distributed cash to voters. As the top panel of Fig. 6.7 shows, a plurality of provincial brokers said that payments were made in the final 24 hours leading up to the vote. In contrast, district brokers seem to have no specific time frame in mind for handing out cash (bottom panel), suggesting that vote buying can occur at any time even *before* the 'quiet period' that begins three days before the vote when no campaigning is permitted. The modal answer is "the night before voting day," which is consistent with

**Fig. 6.7** When did provincial brokers (top panel) and district brokers (bottom panel) hand out cash to voters? (%). Source: My survey of brokers, 30 September–25 October 2014

the conventional wisdom, but this option only had a razor-thin lead. It is a widely held view among candidates and brokers that voters will often vote for the candidate who distributes payment closest to the vote, as the final payment is considered to have a greater impact in voters' memories (a point I return to in a moment). But why was this not the case for district candidates?

Because they occur in smaller constituencies, the intensity of vote buying is greater in the district-level DPRD contests, where candidates can secure a seat with only a few thousand votes. The smaller constituency size allows candidates and their political operatives to gain knowledge about their rivals' tactics. When they learn that rival teams have begun distributing cash, they are typically provoked to follow suit. As argued in Chap. 5, candidates often feel anxious that their supporters will desert them in favour of cashed up rivals, if their loyalty isn't secured by way of a payment. In contrast, in the larger provincial constituencies, it is more difficult to keep tabs on competitors. Overall, however, the big picture remains unchanged: as the elections draw near, the probability of vote buying increases, as I will elaborate in a more detail in Chap. 7. Interestingly, as shown in both the top panel and bottom panel of Fig. 6.7, post-electoral payoffs aren't common in Indonesian elections. This corresponds with the widespread perception among brokers who believed that the majority of voters would vote for the candidate who distribute cash last—just before polling stations are open, so that distributing handouts after the election is wasteful.

### 6.7 How Do Brokers Extract Rents?

From the outset, I have stated that my explanation of vote-buying patterns in Indonesia, which integrates attention to a party-loyalist strategy with emphasis on personal networks, points to a style of vote buying that is prone to rent-seeking behaviour. My study is consistent with the growing interest in the literature on the issues of broker predation and defection (e.g. Aspinall, 2014; Stokes et al., 2013). In my interviews and FGDs with candidates, they raised deep concerns about the unreliability of brokers, especially those who didn't enjoy prior relationships with them. Confronted by the risk of embezzlement and defection by their brokers, candidates therefore constructed their personal campaigns on the basis of relational and personal networks in order to minimise brokers' misbehaviour. As noted earlier in the chapter, however, due to the limited availability of brokers who have personal ties to the candidate, most ended up building 'extended success teams' involving at least some persons with no *direct* personal relationships with the candidate. Drawing from my unique dataset of brokers and candidates and close analysis of campaigns during the 2014 elections, I find at least five pieces of empirical evidence pointing towards agency loss7—which implies that brokers can extract rents from candidates—and broker unreliability.

First, this study found significant evidence of brokers' misbehaviour, especially defection, or at least the potential for it. Overall, 28% of brokers in my survey admitted they were asked to support multiple candidates during the 2014 election. Provincial brokers were more likely to be approached by multiple candidates, especially from the same party. Among those being targeted by multiple candidates, 39% were approached by candidates coming from a different party. Interestingly, district brokers were more likely to be asked by multiple candidates from *different* parties. The survey also asked: "Were those candidates competing at the same legislative level (DPR RI, DPRD I, DPRD II) or a different level?" The modal answer is "different legislative level" (61%), which potentially suggests that even if they ended up working for more than one candidate, they may not have been in direct competition for the same votes. However, 29% of brokers were asked by multiple candidates who competed at the same legislative level and in the same electoral district and were therefore in direct competition. The remaining 8% were at the same legislative level but in different electoral districts. Evidently, district brokers were more likely to be targeted for defection than provincial brokers. There were hundreds of thousands of candidates running for seats in district-level legislatures, simultaneously trying to draw on the same supply of brokers, creating incentives for opportunistic brokers to defect or to split their efforts between more than one candidate.

Second, as implied by the territorially pyramidal structure of success teams used by most candidates (Aspinall, 2014), it is possible that "some of the money they were distributing would go missing, as it had to be passed through several sets of hands… before it reached the voter" (Rohman, 2016: 243). While considering social desirability that might lead brokers to underestimate rent-seeking, this study used an obtrusive measure: "Did you receive any envelopes/cash to distribute to voters from

<sup>7</sup>Chap. 5 conceptually discusses agency problems in the relationships between candidates and brokers and the interactions between brokers and voters in which both problems bring costs to candidates.

the candidate or campaign coordinator?" About 16% said yes; of those, 12% reported diverting resources for their personal benefit. Only 5% of brokers reported receiving staple goods to be distributed to voters; of those, 11% expropriated these in-kind goods. The actual cases may well exceed those figures, but these figures show that the widely held assumption that many brokers engage in extracting rents is well founded.8

Third, I did find a consistent pattern in the dataset showing that brokers have an interest in exaggerating the number of supposedly base voters and even in deceiving their candidates on this issue (see Chap. 5). The survey asked: "In your opinion, are you certain that your neighbours' swinging voting intention can be directed according to campaign team's persuasion if given assistance (envelope/money/gifts, construction of houses of worship, irrigation, roads, etc.)?" Nearly 70% of brokers claimed that they were "very" or "quite certain" they could influence their neighbours' voting decisions by giving them material inducements. However, this response might not simply be a measure of the susceptibility of voters to patronage but also a reflection of rent-seeking behaviour on the part of brokers. When brokers provide high estimates of the effectiveness of their gifts, the more cash they will have to distribute, and the more they can engage in predation. By promoting this expectation, brokers are sending a message to candidates to provide more benefits for them to distribute.

The issue here is that though brokers express confidence in their ability to sway voters with gifts, research on voters themselves provides little evidence to support this view. Recall that in my post-election survey of voters, respondents were asked to assess whether vote buying was acceptable or unacceptable. Among those who thought it was acceptable (40%), a follow-up question was provided: "Will you accept those money or gifts?" The modal answer was that they would take the money but vote based upon their conscience (57.7%), while some answered they would 'not accept.' This evidence confirms the classic problem of vote buying being an uncertain business. Funneling benefits to uncommitted voters is a risky bet because they can behave opportunistically (Diaz-Cayeros et al., 2012: 77). Asked what they would do with voters who received payments but

<sup>8</sup> Indeed, we found quite a small number of brokers admitting rent-seeking behaviours. Note that the population of my survey was brokers who worked for successful candidates who, as discussed above, mostly had had clientelist interactions with them. While acknowledging social desirability bias, it may be that candidates who built success teams on the basis of personal and long-lasting relationships were more likely to win because their brokers were less likely to extract rents.

didn't come to the voting booth, 73% of brokers said they wouldn't do anything. Why, then, did brokers insist on directing benefits to such voters when they couldn't guarantee that the recipients would reciprocate on voting day? By urging their candidates to distribute money or gifts to uncommitted voters, brokers had greater opportunities to extract rents.

Fourth, another common example of shirking was that the candidates' claims that most brokers worked on the basis of systematic voter lists proved to be unfounded. Most candidates say that they instruct groundlevel brokers to draw up lists of potential voters willing to support them, a process known locally by the English-language term, 'by name, by address.' Some candidates I encountered even showed me a stack of their brokers' voter lists. However, it turns out that success team members who compiled voter lists weren't as numerous as candidates expected. Only 47% of provincial brokers admitted that they drew up voter lists and 52% of district brokers did so. If it is the case that most candidates had these lists, why did only about half the brokers reportedly produce them? A candidate from PAN admitted that many candidates were tricked by their brokers, who provided them with fake voter lists. Accordingly, he always quality controlled the lists by conducting spot-checking, hiring a polling firm for this task, and finding that some people whose names were on the lists were already dead (Interview, 22 April 2014). It is also plausible that some brokers worked for multiple candidates at the same time, giving them exactly the same voter lists (Triantini, 2016: 258).

Even among those who provided voter lists, not all of the names on the lists ended up voting for the candidate a broker supported. Some opportunistic brokers intentionally filled their lists with names of individuals whose voting behaviour couldn't be determined, simply as a way to generate profits. Recall that most candidates rely on these lists as a basis for delivering cash payments. According to Triantini (2016: 250), in the case of Blora, Central Java, base-level brokers locally called *sabet* "used rather slipshod methods in drawing up the lists, writing down people's names without knowing anything about their preferences, or without inquiring about which among them had been contacted by other teams." My brokers' survey found similar evidence. About 22% of brokers reported putting some or most of the names on their list without first consulting the voters.

Therefore, candidates expect that there will be a gap between the number of the names on the list to whom they distribute payments and the final number of votes they receive. They typically use a term derived from English, 'margin error,' (sic.) for describing this discrepancy, blaming 'extended brokers,' whose primary motivations are largely the pursuit of material rewards rather than long-term personal relationships with them. In my interviews, when candidates talked about this 'margin error' they typically referred to two sources of failure: (1) unreliability of voters who receive cash but who don't feel bound to repay with their votes and (2) failings of brokers, either due to rent-extraction by brokers or lack of capacity to identify and distribute benefits to voters who would reciprocate with votes. For their part, brokers typically point their finger at the transactional nature of voters who might vote for more 'generous' rivals.

Confronted by pervasive broker predation, some candidates claim to keep a tight rein on their success team members by doing multiple checks of the names on the lists. Realizing that the determinant of electoral success is the strength and reliability of brokerage networks, Saan Musthafa of the Democratic Party hired a Jakarta-based polling organisation to survey the effectiveness of his teams including by randomly sampling the people on the lists to check whether brokers had really talked to them (Interview, 25 April 2014). As I elaborate in more detail later in the chapter, however, candidates in general had very few instruments to monitor their brokers. Having seen that they were mostly unable to build tight and strong monitoring of their teams, candidates were responding by trying to pick brokers on the basis of intimate and personal contacts. However, due to the limited number of those who were emotionally and personally close to candidates, it was hard to avoid recruiting at least some brokers who lacked such ties.

Lastly, aside from the rent-seeking behaviour, another form of principalagency problems between candidates and brokers arose because some brokers actually didn't know their supposed clients well. My argument here is different from what Stokes and her colleagues' (2013) concluded—that brokers are necessary for clientelist exchange because they know their clients intimately. They argue (2013: 96) that "brokers are indeed involved in long-lived interactions with their neighbours and clients, interactions which—in the brokers' view—give them privileged information about the preferences and behaviours of individual voters." They offer detailed evidence from their broker survey in Argentina suggesting that nearly 80% of brokers claimed they would know when their neighbours, with whom they have a lot of dealings, voted against the candidate they were 'supposed' to support. Although this measure is likely to be distorted by social desirability bias, it is a simple measure of brokers' confidence in their ability to observe their clients' political preferences and actions.

Using a similarly worded question, my survey found a somewhat different picture in Indonesia. Indeed, 57% of brokers claimed to be able to infer whether their neighbours voted against their candidate.9 But the proportion of those who said they couldn't ascertain their clients' preferences was more than double the figure Stokes and her colleagues found in Argentina. As a follow-up, my broker survey asked those ground-level operatives, who claimed to know when a neighbour voted against a candidate they supported, how they did so? How could they work this out? The most frequent answer was derived from their day-to-day interactions with clients, enabling them to 'draw inferences from attitudes or affect' (37%) such as being socially awkward, or through mannerisms or change of attitude, a look or gesture, or trying to avoid them. Others said they found out by 'asking around through direct communications about how a neighbour discusses any particular party/candidate' (31%), whereas, others said that they had been simply informed by their neighbour (5%), or based on information provided by other neighbours who stayed loyal (5%), or through closely monitoring their data collection or internal surveys (3%), or through the display of campaign posters or propaganda tools for another candidate in the voter's home (3%).

But the key point is that in Indonesia there are widespread perceptions of ballot secrecy as evidenced in the relatively large number of brokers who recognise the difficulties they have in inferring their neighbours' vote choices. These findings are consistent with a widespread, almost 'doctrinal' belief that the country's elections must be 'direct, general, free, and confidential' (*luber*) that runs deep among voters. In my post-election survey of voters, we asked respondents whether, despite the principle that elections are confidential, influential people could discover how they voted. Only 19.7% responded that this was "somewhat likely" or "very likely." The majority of respondents confirm the conventional wisdom that the secrecy of the ballot is difficult to violate, and even powerful persons in the neighbourhood wouldn't be able to find out how an individual voted.

In sum, brokers have both strong incentives and opportunities to shirk. The problems of predation and defection are especially severe in the case

<sup>9</sup>The question reads: "If you had a good relationship with your neighbours, when they voted for the candidate/party you didn't support, would you know about it?"

of extended success teams whose commitments to the candidate are more transactional and whose members lack prior contacts with the candidate. As I explore in more detail in the following pages, many candidates pour out cash handouts en masse, but they ironically invest little effort in monitoring and disciplining brokers.

# 6.8 How Stringent Was Candidates' Monitoring of Brokers and Voters?

As alluded earlier, candidates' capacity to monitor broker performance is weak. Surprisingly, my broker survey reveals that only 10% of the sampled brokers admitted that their actions were often monitored by their superiors in the team or campaign coordinators; 18% acknowledged that their campaign work was sometimes being monitored and a large majority felt they weren't monitored at all. Candidates are also vulnerable to rent-seeking behaviour by brokers because there are few punishment mechanisms for brokers who don't meet their target numbers of votes. A total of 56% of brokers admitted to failing to meet their targets. When asked a hypothetical question about the consequence if a broker fell far short of his or her target,10 only 7% of respondents said the failing broker would receive any negative treatment from the campaign coordinator or candidate. Among those who responded affirmatively, a follow-up question was asked about the form of negative treatment: the answers were that the failing broker would be subjected to verbal abuse (30%), not invited to join success teams in the future (28%), cut out of future pork-barrel projects (20%), or not receive the money or goods that had been promised (9%). All of this suggests that most candidates are unable to build stringent mechanisms to monitor or discipline their brokers.

Not only did candidates have few instruments to enforce discipline on their brokers, the brokers themselves had even fewer mechanisms for monitoring their voters to ensure that they kept their bargain to vote for the candidate. Asked what they could do when voters received gifts but voted for another candidate, the overwhelming majority of brokers said they couldn't do anything. Similarly, brokers couldn't do anything with voters who took the money but didn't even attend the polling station

10The question reads: "If there is a broker who didn't meet the target (for example, he/ she promised to generate 20 votes, but the candidate finally got only 2 votes at the polling station/village), did the broker receive certain negative treatments from the campaign coordinator or candidates concerned?"


**Table 6.6** What would you do when a recipient didn't come to the voting booth?

Source: My survey of brokers, 30 September–25 October 2014

(Table 6.6). Only a few brokers reported they would scold or threaten those who just took the money but voted for other candidates.

Why did brokers seem to be pessimistic about enforcing their deal with voters? The answer primarily lies in the fact, already touched upon earlier in the chapter, that elections are free, and that ballot secrecy is strongly enforced in Indonesia, incentivising voters to behave opportunistically. My voters' survey reports very minimal rates of intimidation or violence from political campaigns, suggesting that both brokers and voters who were unable to comply with the bargain would rarely be targets of violence. An overwhelming majority (84%) considered the 2014 election as 'fair and free' based on the exit polls run by Indikator, to which I am affiliated in which 1928 respondents were interviewed face to face immediately after casting their votes.11 There is widespread confidence that vote choices can be kept secret from politicians and their success teams. My postelection survey of voters asked respondents whether politicians could guess how they voted. A total of 74.8% of the sample responded this was "not at all likely" or "not very likely," while only 13.2% said it was "somewhat likely" or "very likely."

Likewise, when asked to name who else might be aware of how they voted, though they were allowed to provide multiple responses, only 2.3% of the sample responded that party or a candidate brokers could find out how they voted (Fig. 6.8). The modal answer is husband or wife (58.8%) then followed by relatives (14%), suggesting that many voters don't discuss their electoral choices with outsiders or, sometimes, even within the

<sup>11</sup>Available at http://indikator.co.id/uploads/20140411204045.Hasil\_EP\_Pileg\_2014\_ Update.pdf. Accessed 5 October 2016.

**Fig. 6.8** Who else do you think is aware of how you voted? (%). Source: My post-election survey, 22–26 April 2014

family. One senior PKB politician colourfully recounted a story of his neighbour from his hometown in Central Java who, despite receiving a lot of benefits from him, shamelessly requested additional compensation in return for his vote; the politician ironically noted, "Still, which candidate he finally cast his vote for was a mystery to me" (Interview, 20 April 2014). With echoes of earlier work by Aspinall and his collaborators (2015), much of what I see in Indonesia provides little support for Stokes' (2005) theory of "perverse accountability." She argues that instead of politicians being accountable to voters, where vote-buying transactions occur, voters are held accountable for their vote through direct and indirect coercion. This wasn't the case in Indonesia. Even those who take rewards don't lose their power to hold politicians accountable because political machines are unable to infer voters' behaviours; voters can still take money from a candidate and vote against him or her.

#### 6.9 Conclusion

In this chapter, I have outlined that, unlike the comparative literature on vote buying which conventionally assumes that political parties play a central role in the distribution of material inducements, political parties in Indonesia have a relatively marginal role in electoral campaigns. Under Indonesia's open-list PR system, the influence of party organisation is limited, and it is individual candidates who have more prominence in electioneering. Additionally, in settings like Indonesia where partisan ties are weak and where the personalised nature of voting requires co-partisans to compete against each other, candidates rely on more personalised rather than strictly party-based relationships. As a result, personal networks and non-party brokers play the critical role in grassroots campaigning.

At the outset of the chapter, I demonstrated that personal networks not only structured vote buying but also had significant impacts on (1) broker recruitment, (2) success team structures, and (3) targeting strategies. Personal networks shape how candidates recruit success team members and determine the structure of vote brokerage. More than two-thirds of sampled brokers knew the candidate personally prior to joining his/her success team. There are two logics underpinning this strategy. First is to minimise the problem of broker loyalty. Those who have personal connections with candidates are less likely to shirk. Second, from a purely strategic perspective, such brokers are more likely to campaign on the candidate's track record. The centrality of personal networks was also visible in targeting strategies. Brokers prioritised household members, close friends, and neighbours when collecting votes. When candidates and brokers started to exhaust the supply of voters who have personal connections to them, they move further afield in the search for both brokers and voters. Accordingly, a substantial amount of leakage—which occurs when people receive payment yet don't reciprocate with votes—occurred. Given that many beneficiaries were selected on the basis of personal networks, which were loosely connected to ideological proximity to the party or candidate, the potential for such slippage was great.

It is increasingly clear that buying votes was a risky game for candidates. Given their lack of mechanisms to monitor their brokers and the reality that many of them used 'extended success teams,' relying on at least some brokers with whom they lacked direct personal connections and who were more concerned with material payoffs, their vote-buying attempts were vulnerable to rent extraction by brokers. Fine-grained evidence from my unique broker survey combined with qualitative work demonstrates that rent-seeking behaviours among brokers were common in the 2014 legislative election. Nonetheless, candidates were quite relaxed about their brokers failing to meet vote targets, at least if they were successful in securing enough personal votes to win. Having realised that such rent-seeking behaviours are unavoidable and there are so many difficulties in enforcing the vote-buying contracts, candidates are tolerant if their brokers fall short, as long as they produce victory in the context of Indonesia's highly competitive electoral landscape. Finally, despite the personal networks helping to provide a mechanism to structure vote buying in Indonesia, they cannot escape from the problem of agency loss between candidates and brokers that is integral to electoral clientelism in many contexts.

There remains one major puzzle, however. If vote-buying efforts are so vulnerable to broker predation, as shown in this chapter, and the targeting is so misdirected, as already discussed in Chap. 5, why do candidates invest so heavily in it? The next chapter provides an answer to this question.

#### Bibliography

Books, Journal Articles, and Unpublished Papers


**Open Access** This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

# Does Vote Buying Affect Voting Behaviour? Chasing Winning Margins and the Prisoner's Dilemma

Making vote buying work is extraordinarily difficult, especially in the presence of ballot secrecy. Yet vote buying is extremely widespread in Indonesian electoral politics. Furthermore, vote-buying efforts face serious problems in their targeting: as we have seen, political actors try to 'buy' the votes of those who look like their loyal supporters, yet in practice end up giving most payments to other voters. The problems get worse as a result of rent-seeking behaviours by brokers, which increase the inefficiency of vote buying. All this leads to an obvious puzzle: if such electoral handouts are so misdirected and create strong incentives for brokers to extract rents, why do candidates invest so much money and goods in them? The answer must be found at least partly in the effect of vote buying on electoral behaviour. How effective is vote buying in actually winning votes? Surely it must have some effect in order for candidates to pursue it?

Little research has been conducted in Indonesia to measure the influence of vote buying on voting behaviour. In the beginning part of this chapter, I endeavour to quantify the impact of vote buying on voter turnout and the vote shares won by candidates. I find that handouts actually produce a sizeable turnout or higher vote share for the distributing candidate. However, I also find payments influences decisively the votes of 'only' about 10% of people who receive them. In this seemingly low number lies the key to the attractiveness of a vote-buying strategy. The proportion of recipients who admit their choice is influenced by material inducements is more than enough to constitute a small margin of victory for most candidates.

However, chasing small margins isn't the whole answer for why candidates buy votes. There are many countries and elections where the vote margins are small, and still candidates don't engage in vote buying. Accordingly, I would argue that even if the returns are so poor, and even if candidates aren't running in a close battle, they still engage in buying votes in the presence of rampant handout distribution by their competitors towards the responsive voters in the run-up to the election. Even so, it is widely assumed among political machines that although vote buying may be an *objectively* inefficient strategy for generating votes, it is still more efficient than all other viable alternatives.

### 7.1 Effect on Voter Turnout

As its name implies, vote buying is often defined in the literature as a direct market transaction where voters provide their vote in return for money or gifts (Guardado and Wantchekon, 2014). Brusco and her collaborators (2004: 67), for instance, define vote buying "as the proffering to voters of cash or (more commonly) minor consumption goods by political parties, in office or in opposition, in exchange for the recipient's vote." In a similar vein, Finan and Schechter (2012: 864) view vote buying as "[offered] goods to specific individuals before an election in exchange for their votes." Given that vote buyers often don't explicitly demand a vote in exchange for their payment, Aspinall and Sukmajati (2016: 20) slightly modify the definition as "the systematic distribution of cash payments and/or goods to voters in the few days leading up to the election with the implicit expectation that recipients will repay with their vote." Similarly, Kramon (2009: 4) defines vote buying as "the distribution of particularistic or private material benefits with the expectation of political support." All of these definitions assume that paying boosts voter turnout and/or the vote share of the paying candidate or party (Guardado and Wantchekon, 2014: 2).

In practice, however, vote buying is an uncertain business. How do vote buyers ensure that the vote that is being sold is actually provided, especially in the presence of ballot secrecy? As previously presented in Chap. 6, most electoral payoffs in Indonesia are provided before the elections. If this is the case, it is possible for the recipients to behave opportunistically: to take the money and run. But if a candidate was to promise to provide cash only after the election, voters would likely suspect that the candidate would break his/her promise and not deliver (Baldwin, 2016: 67). If vote buying instead takes place before the election, with voters accepting payment but still being allowed to vote based upon their conscience, as Kramon (2009: 2) has questioned, why might vote buying have an influence on voting behaviour?

The effectiveness of vote buying clearly relates to the party loyalist versus swing targeting debate in the literature. If we stick with the party loyalist argument, which views vote buying as turnout buying in which parties or candidates target voters who are already inclined to support them, the payment looks more like the mobilisation of passive supporters to come to the voting booth rather than 'buying' the vote of an indifferent voter. In contrast, the swing-voter hypothesis implies that the payment really does act to purchase the support of an uncommitted voter. In this case, monitoring whether recipients turn out at the polls is less of an issue for the distributing party than is monitoring vote choice (Nichter, 2008). In the turnout-buying model, if a passive supporter bothers to turn out, the givers can be confident that the voter will choose them. This sharply contrasts with the model presented by the swing-voter school,1 under which even if the recipient shows up at the polls, the vote buyer will still have no guarantee whether that person votes for the buyer or some other candidate.

Regardless of such challenges, I argue that politicians still have incentives to pursue vote buying because the evidence suggests it seems to influence voting behaviour in Indonesia. Let us first discuss its impact on voter turnout. My survey of voters conducted immediately after the 2014 parliamentary election allows me to examine the effect of vote buying on turnout. While the measure of vote buying has been already discussed in Chap. 2, the wording for the question on turnout was: "When discussing the election with others, we found many people couldn't vote because they were far from home, sick, or didn't have the time or other reasons. What about you? Did you vote during the last legislative election on April 9, 2014?" Given that social desirability bias might induce the respondents to overstate their voting histories, I weighted the reported turnout by using the official turnout rate according to Indonesia's General Election Commission (KPU). The commission reported that national turnout for the 2014 legislative election was about 75% (Table 7.1).2

To test how voter turnout is affected by electoral handouts, I ran a cross-tabulation followed by chi-square to determine whether or not a null

<sup>1</sup>Abstract Politics. (2008) http://abstractpolitics.com/2008/05/vote-buying-or-turnoutbuying/.

<sup>2</sup> See http://www.kpu.go.id/koleksigambar/Partisipasi\_Pemilih\_pada\_Pemilu\_2014\_ Studi\_Penjajakan.pdf.


**Table 7.1** Cross-tabulation of a respondent's reported turnout and their likelihood of being offered vote buying (%)

Source: My post-election survey, 22–26 April 2014

hypothesis can be rejected. The 'don't know' or 'refuse to answer' options weren't included in the analysis. My primary interest was to examine the hypothesis that respondents who were exposed to vote buying were more likely to participate in the 2014 elections. The cross-tabs seem to support the notion that electoral handouts are quite effective at producing a higher turnout. About 81% of the respondents who received a pre-electoral benefit showed up at the polls, compared to 74% of those who didn't receive offers of vote buying. Turning to the measures of strength and significance, the Pearson chi-squared value is 0.017, meaning that it is below the *p*-level of 0.05, thus making it significant. Although vote buying and voter turnout do have a statistically significant association, it is also reasonable to argue that vote buyers were targeting voters they believed or knew were more likely to vote. As argued in Chap. 2, this notion is in line with the evidence that candidates and brokers tend to target more heavily that group of voters for whom vote buying is an acceptable practice, which in turn increases their likelihood to vote if given rewards.

Overall, the findings provide initial suggestive evidence that vote buying may have a significant impact on turnout, but we cannot be sure about the direction of causation: it may be that success teams target for payment voters who they identify as being more likely to turn out.

### 7.2 Effect on Vote Choice

Having reviewed the effect of vote buying on voter turnout, we are now in a position to test its influence on voting choice. I return to data from my large nationally representative survey conducted immediately after the


**Table 7.2** Relative influence of vote buying between 2009 and 2014 (%)

Source: My post-election survey, 22–26 April 2014

2014 legislative election. The data allow me to quantify the effects of vote buying.

Table 7.2 provides a sense of comparison between reported vote buying and its effects on vote choice3 in the 2009 and 2014 legislative elections based on the direct individual questions. Recall that the question about the influence of vote buying was only asked to those who admitted being offered a reward. Of those subjects (25%) who reported being targeted in 2014, around 41.8% admitted that the handouts were effective at influencing their vote. Overall, then, my study finds evidence that vote buying produces electoral support for the distributing party or candidate. However, its effect seemed to be 'limited,' amounting to 'only' 10.2% of the total electorate.

#### 7.3 Chasing a Margin of Victory

At first glance, the estimate I have come up with for the effect of vote buying may appear small, since the data showed that only a relatively small proportion of recipients of payments reciprocate with votes. These results present a puzzle. If this is true that vote buying yields minor results, why do politicians do it? If the votes of only a small proportion of those to whom they deliver payments are swayed, why do they persist? Note that vote buying isn't an easy task. The problems of broker predation and misdirected targeting already make vote buying tremendously inefficient, as discussed in earlier chapters. Yet on top of these problems, it is also, overall, ineffective at influencing vote choices. Under such circumstances, why do candidates invest large amounts in gifts to voters?

The answer is found in the high electoral uncertainty regarding candidates' personal prospects of victory. Although, the effect of vote buying

<sup>3</sup> If a respondent gave an affirmative reply when asked about vote-buying offers, I asked a follow-up question: "Did the gifts have an influence on your vote?"

on electoral outcomes looks insignificant, a minor shift in votes can make a huge difference for a candidate. It can be the difference between winning and losing in a competitive election. Candidates have reason to invest in vote buying because they are usually chasing a narrow winning margin.

#### *7.3.1 Open-List PR and Electoral Competitiveness*

In order to substantiate this argument, I first establish the extent to which electoral competitiveness affects candidate behaviour. As discussed in Chap. 1, since the introduction of a fully open-list system in 2009, legislative elections have been extremely competitive. Note that in Indonesia's 2014 legislative elections, there were 6608 candidates distributed across 12 national parties running for the 560 seats in the House of Representatives. Therefore, the average level of competitiveness was 11.8 candidates per seat. As discussed in Chap. 1, in order to determine the winning candidate according to the open-list system, each party that successfully secures a seat (or seats) must allocate it (or them) to whichever of its candidates obtained the most votes. If there is only one seat for the party, the winner takes all. The open-list system has thus produced a pattern of 'ground war' electioneering in which candidates from the same party engage in intense campaigning for personal votes (Aspinall et al., 2017: 12). PDI-P's Richard Sualang, for instance, recalled that during the 2014 campaigns one candidate from a different party approached him to release his voter lists (the lists which, as we have seen, many candidates use to determine to whom they will deliver cash payments). If he was willing to hand over the lists, this external rival promised he would use them to ensure that he wasn't targeting Sualang's base voters, and to ensure that he was instead distributing resources to his own constituents in order to outspend his co-partisans (Interview, 26 April 2014).

As a result of this situation, legislative elections in Indonesia have become zero-sum games. One striking example is a close battle for a seat in the provincial legislature in the electoral district 5, Special Region of Yogyakarta. PDI-P was declared to be the winner and received two seats. Koeswanto won the first seat by a comfortable margin. However, there was great uncertainty about which candidate would secure the second seat because the results were so close. With 99% of the vote counted, it was still unclear who would win. Eventually, the final count gave incumbent candidate Gimmy Rusdin victory by a single vote, meaning that the vote margin was essentially zero. He won the race dramatically with a total of 9417 votes, while his internal rival, Listiani Warih Wulandari, secured 9416 votes.4

Under open ballot systems, candidates have clear incentives to compete against internal party rivals rather than focusing their competition against candidates from other parties. Whether based on past voting records or a strong belief that each party has its own constituency, they are usually able to predict how many seats each party will win in a given electoral district (Ibrahim, 2016), or at least there is relatively little uncertainty regarding the distribution of seats among parties. But they suffer from a high degree of uncertainty regarding which individual candidate will win. This is particularly the case when there is no candidate who is widely favoured to win in a particular district. Even the presence of a very popular candidate doesn't necessarily lower the level of uncertainty. In the electoral district Central Java V, it is almost impossible for PDI-P candidates to defeat the incumbent, Puan Maharani, the daughter of party matriarch Megawati Soekarnoputri. But although Puan regularly wins one seat, other candidates from the same party still have a chance of getting elected to the additional seats in the constituency.

The dominant narrative among candidates is that open-list PR systems offer a degree of hope of electoral success to all candidates as individuals, and that their electoral fate therefore depends heavily on their own efforts. Note that in order to gain a seat, candidates first need to make sure their party reaches the 2014 national threshold for parliamentary representation (3.5% of the national vote) and reaches the quota required to gain at least one seat in their electoral district (this is the total number of valid votes cast in the electoral district divided by the total number of seats). The total vote for the party and its individual candidates is therefore important. Assuming a party gains one seat in a district, since that seat goes to the candidate on the party list who obtains the most personal votes, most elected candidates are helped by voters who vote either for the party only or for other candidates from the party. In short, the main challenge for candidates is to be ranked above their co-partisans. Hence, most candidates approach an election feeling they need to figure out how many personal votes they need to win, how close the race will be, who their main internal rivals are, their relative areas of strength, and so on.

<sup>4</sup>E-Parlemen DPRD DIY, "Daftar Caleg Terpilih DPRD DIY Periode 2014–2019," 25 April 2014.

The literature on electoral mobilisation shows that politicians act strategically. If they have little chance of getting elected, they don't invest large amounts of resources in personal campaigns. Likewise, if they have a reasonable chance of winning a seat, they will make more of an effort to compete (Milazzo and Karp, 2013). Selb and Lutz's important study (2015) found that the level of competitiveness isn't only determined by actual election results but also by candidates' self-perceived competitiveness. Candidates facing a narrow loss or narrow victory are likely to spend heavily in search of personal votes to outdo co-partisans, generating a cycle of competition which results in even more competitive elections.

This setting is clearly relevant to my inquiry into vote buying. The great uncertainty surrounding electoral outcomes in places like Indonesia creates incentives for candidates to pursue vote buying to maximise the chances of winning (Kitschelt and Wilkinson, 2007; van de Walle, 2007). The literature has long stated that vote buying tends to be higher in constituencies where elections are highly contested.5 The rationale is simple: in an environment in which a relatively small percentage of the votes can change candidates' electoral fortune, their propensity to engage in vote buying increases. In this regard, vote buying is a means of reducing electoral uncertainty (Jensen and Justesen, 2014). Electoral uncertainty is in fact the defining feature of electoral competitiveness (Przeworski, 1986; Schedler, 2013). The more uncertain the outcome of an election, the more competitive it is (Blais and Lago, 2009: 95; Franklin, 2004: 56–57). It follows that under an electoral system which creates competition for personal votes, candidates' uncertainty regarding the probability of winning will make them consider vote buying as a way to chase even a narrow margin of victory.

In this study, electoral competitiveness is measured at the national level and is operationalised as margin of victory. Using official election statistics, this chapter presents two different measures of the margin of victory—one taken as a percentage of overall valid votes in the electoral district and one as a percentage of the valid votes per party in the district. Note that this may be problematic when comparing to the effectiveness of vote buying: a single vote buyer doesn't distribute payments either to all voters in the electoral district, or even to all who support his or her party, but to a sig-

<sup>5</sup> Indeed, there is an issue of reverse causation here, which I will address in the final part of this chapter, whether candidates are more inclined to buy votes in more competitive districts or whether more vote buying makes districts more competitive.

nificantly lower number. But, at least, the latter (i.e. a percentage of valid votes cast for a party in an electoral district) is a better measure for assessing the margins needed by individual candidates to win by distributing cash (remembering that no candidate will distribute cash to all the party's voters in an electoral district, let alone to all voters). Hence, my analytical focus is on victory margin as a percentage of the valid votes cast for each party in the district.

In this section, however, I start determining average margin of victory as a percentage of all valid votes cast in the constituency, simply to provide a broader picture of how competitive parliamentary elections in 2014 were in each constituency. However, given that under open ballot systems electoral competition takes the form of intraparty competition, and considering that under the system, the primary focus of candidates (the main vote buyers) is getting themselves a seat before their intraparty competitors, the level of competitiveness should be closely examined within political parties. Accordingly, after establishing the difference between the vote share of the lowest-placed winner and the highest-placed losing candidate from the same party in any electoral district, I will discuss the primary interest of this study: margin of victory as percentage of votes cast for a particular party in the constituency in the following section. Then, as I elaborate later in this chapter, in order to test the relationship between vote buying and such electoral competitiveness—measured as the margin of victory—I merge the actual election results with pre-electoral district surveys to gain ex-ante information on the ubiquity of vote buying at the constituency level.

Before proceeding, I present descriptive findings about the closeness of electoral results in each electoral district using victory margins of individual candidates over their party-list rivals as the primary measure of electoral competitiveness, with those victory margins calculated as a percentage of the total valid votes cast in the constituency. Figure 7.1 showing the average margins of victory in each national parliamentary constituency is simple: the larger they are, the less competitive is the electoral district. Because the quota—determined by population size—varies considerably across electoral districts, the average margin of victory in each district is then divided by the total number of valid votes cast in that constituency to produce the percentages in Fig. 7.1. To make it simple, I categorise the results into four broad groups. The first group is ultra-close contests those with a victory margin of less than one-half of 1%. In 2014, the smallest margins were seen in West Java XI and East Java III—its precise margin

was actually 0.47%, rounded up to 0.5% as appeared in Fig. 7.1. The second is very close races, those with victory margins of between a half of 1% and 2%.6 The majority of constituencies (48 out of 77 electoral districts across Indonesia) belong to this group, confirming the hypothesis that the general pattern has been extremely competitive under open-list PR system. It is also worth noticing that in the above graph, most of those very close contests were located in the densely populated islands of Java and Sumatra with some of them decided by even less than 1% of the total polled votes.

The third group is close contests—decided by an average margin of between 2.1% and 4% in 2014. As shown in Fig. 7.1, 17 electoral districts are in this category. A last category is uncompetitive electoral districts which had an average winning margin of over 4%. Interestingly, only a handful of constituencies were decided by big margins. The least competitive constituency in 2014 was Gorontalo, where winning candidates had an average 11.8% margin of victory, followed by Bangka Belitung which had a fairly high margin of 6.9% and Riau Islands and West Papua which equally had a margin of 6.6%. These uncompetitive races all occurred in electoral districts with low magnitude with only three seats available in each constituency. This finding parallels Carey and Shugart's (1995: 431) argument that where district magnitude is higher, incentives to cultivate personal votes increase. Under open ballot systems, the higher a district magnitude, the more co-partisan competitors enter the race, resulting in more competitive elections as a result of increasing intraparty competition. The reverse is also true.

Overall, 69 out of 77 constituencies were decided by slim margins of less than 4%, calculated as a proportion of all votes cast in the electoral district, suggesting that the level of competitiveness in the constituencies was extremely high. As noted above, such constituency-level campaigns were systematically associated with intraparty competition in which candidates from the same party were busy fighting against their co-partisans. Such fierce competition between individual candidates for personal votes helps explain why candidates pursue vote buying, despite its seemingly small effects on vote choice. Recall that the effect of vote buying on voting decisions was up to 10% of the electorate. There were around 187 million voters in Indonesia's 2014 legislative election. Even if we use such an estimate, the 10.2% effect would mean an estimated 19 million voters

<sup>6</sup> I adapt the first two categories from Ray Christensen and Kyle Colvin (2007).

nationwide admitted that receiving money and gifts can be a crucial factor influencing their voting decisions.

Note that the estimates of vote-buying effectiveness are calculated as a percentage of those who received money from a candidate rather than the electorate. Given that 'only' 25% of respondents were exposed to vote buying (based on the direct measure), and 41.8% of the recipients were influenced, in total numbers, vote buying had an influence on 10.2%. The effect would likely have been higher if the candidates were able to distribute payments to more than a quarter of the electorate. Given that among those being targeted, 41.8% admitted that the handouts were effective at influencing their vote, it can be inferred the more the number of envelopes candidates distribute, the more likely they are to generate higher vote share. Let's say that an electoral district has 1,000,000 valid votes, and an average margin of victory of 4%—4% is 40,000 votes. If a candidate gave cash to 100,000 voters, he/she would generate 41,800 votes (41.8% of the 100,000 recipients), assuming he/she is the only candidate who engages in a vote-buying operation. This amount would be more than enough to explain the victory.

It is more complicated in practice, though. As I will explain in the next section, given that there are multiple candidates competing to purchase the votes, or even avidly bidding up the price of votes to defeat rivals (Aspinall et al., 2017), and the evidence that a significant number of the electorate received multiple payments, it would be difficult for candidates to assess how successful their vote-buying efforts was in generating votes. Therefore, many wealthy, serious candidates often double their efforts at vote buying in the hope of reducing uncertainty with regard to the election outcomes and maximising their individual chances of success.

This helps to solve the above puzzle of why politicians insist on spending money on vote buying in legislative elections, despite the fact that it is a strategy that would seem to fail to yield full effects. Despite vote buying being vulnerable to broker predation and the recipients not always repaying with votes, politicians believe that minor changes in voter support whether by buying votes or other electoral strategies—can make a difference to the outcome.7 Overall, this is in line with previous works (e.g. Jensen and Justesen, 2014), suggesting that vote buying is a key instrument for parties or candidates to create winning margins.

<sup>7</sup>For further discussion on the impact of competitive electoral settings where minor shifts in vote shares can change electoral results, see Milazzo and Karp (2013).

#### *7.3.2 Intense Intraparty Competition*

Having established the average margin of victory in each constituency, we are now in a position to provide the average victory margin in each *political party*. As noted above, due to Indonesia's open-list PR system, which incentivises intraparty competition, this measure is a better indicator for assessing the competitiveness that candidates care most about. To arrive at the figure, the average margin of victory is calculated as the lowest winner's votes minus the losing runner-up's votes from the same party divided by the total number of valid votes polled for that party in the electoral district. Overall, the size of the winning candidate's victory within a political party varies significantly. For instance, the winning party in 2014, PDI-P, won 109 of the 560 seats in the national legislature, with the victory margins of these candidates over their nearest-placed PDI-P competitors ranging from 0.1% to 67.2%. At the high end, PDI-P's Jimmy Demianus Ijie of West Papua defeated his nearest co-partisan by the widest margin. At the low end, PDI-P's Wiryanti Sukamdani of Jakarta I seat scraped through with a small margin of 441 votes or equal to 0.1%, the lowest margin among all PDI-P winning candidates. A President Director and CEO of PT Sahid International Hotel, and a daughter of one of the richest men in Indonesia, Wiryanti not only defeated her nearest party rival Abadi Hutagalung, she also successfully unseated the incumbent candidate Adang Ruchiatna. As alluded to above, PDI-P candidates also fought in a close race to compete for the third seat in one of the party strongholds in Central Java V. The promising young professional Darmawan Prasodjo lost to Rahmad Handoyo by a margin of 485 votes or 0.1%. Among electoral districts won by PDI-P, 27 constituencies witnessed victory with a margin of less than 4%. Among others, East Nusa Tenggara I saw the closest fight as Honing Sanny, who polled 49,287 votes, beat an intellectualturned-politician widely known as Megawati's surrogate Andreas Pareira, who received 49,089 votes; West Java VII was among the most closely contested constituencies where four PDI-P candidates had a close fight to compete an additional seat received by the party.

Similarly, politicians from Golkar, the second-placed party in the 2014 legislative election, were forced to compete in very tight races against copartisan rivals in many electoral districts across Indonesia. Of the 91 seats the party won, 26 seats were close victories in which the winning candidates needed a margin of less than 4% to topple their internal competitor. In West Java XI, four candidates from Golkar initially had a chance of winning an additional seat. Ultimately, Zacky Siradj took home the prize by the lowest margin of 0.2%. Likewise, in Central Java IV, Endang Maria Astuti, who occupied a low rank on the party list unexpectedly defeated with a victory margin of 0.3%—high-profile names from her own party, including the sitting candidate Hajriyanto Thohari, former Deputy Chairman of the People's Consultative Assembly (MPR), who ended in the third position. Golkar politicians from the outer islands also witnessed tight wins, including Indro Hananto who defeated his nearest party rival by a 0.3% victory margin in South Kalimantan I, Syamsul Bachri who won with a 0.6% margin in South Sulawesi II, and M. Lutfi who retained his constituency in West Nusa Tenggara, defeating his nearest rival by a margin of 0.9%.

The third-placed party in 2014, Gerindra, also experienced highintensity campaigns among its candidates. Fourteen out of the 73 seats Prabowo's party won were decided by a margin of less than 4%, with some won with a margin of less than 1%. The striking example was Martin Hutabarat, who almost lost his seat in North Sumatra III to his party rival Sortaman Saragih. Only 27 votes separated the winning Martin from the losing candidate, meaning that the margin was basically zero. This was the lowest victory margin not only in that electoral district but also among all Gerindra winning candidates. Likewise, Gerindra's Dairul suffered defeat at the hands of his internal party competitor, H. Anda in Banten I constituency with a small margin of 332 votes. In West Kalimantan, Katherine Oendoen defeated her party rivals by a slim margin of 0.3%, including Deputy Party Leader Arief Poyuono.

In a similar vein, a high degree of intraparty competition happened among the Democratic Party's candidates in 2014. A series of high-profile corruption scandals implicating its party executives (Aspinall et al., 2015a) forced its candidates not to rely on party branding but instead on their individual efforts, intensifying intraparty competition. Almost half of the 61 seats the party won in 2014 were decided by a margin of less than 5%. Dramatically, among these lowest-margin wins, two seats saw victory with winner-loser differences of almost 0%. Salim Mengga retained his constituency in West Sulawesi with a small margin of 25 votes after defeating the closest rival, Sulfia Suhardi. This was the lowest winning margin across national DPR constituencies in Indonesia and across the winning candidates in all political parties. Similarly, Ikhsan Modjo, an economist-turned politician and party leader's ally, was surprisingly defeated by a notorious local politician Ayub Khan with a margin of 57 votes in East Java IV.

Smaller parties also deserve to be mentioned. In Papua, the internal political race within NasDem was heated as several high-profile candidates clashed with each other. Three candidates had a chance of winning, but Sulaiman Hamzah ultimately won the seat, defeating former two-term governor of Papua, Barnabas Suebu, with a small margin of 1%. In general, at least a quarter of NasDem's victories in 2014 were closely contested within the party list. A high degree of uncertainty and intense intraparty competitions can produce what Christensen and Colvin (2007) termed 'election-night corruption' in which one candidate ends up engineering sufficient votes to defeat the nearest rival by a slim margin. For instance, Hanura's top-ranked politician Erik Wardhana was first announced to hold a narrow lead over his co-partisan rival Djoni Rolindrawan in West Java III. The initial vote tallies showed that the sitting candidate Erik would retain his constituency by a margin of 0.9%. Djoni refused to concede, however, and reported Erik to the Elections Supervisory Agency (BAWASLU) and The Election Organisation Ethics Council (DKPP) for allegedly manufacturing votes to win the election. Having proved such fraud, both bodies recommended the General Election Commission (KPU) revise the vote tallies, and Djoni was then declared the winner (see DKPP's Putusan No 30 Tahun 2014; *Media Indonesia*, 22 September 2015).

Zero-sum intraparty campaigns also appeared among candidates running with Islamic political parties. Among the victories with the smallest margins in PKS, for instance, three seats had a margin of less than 1%. Of the three, West Java V witnessed the closest fight as PKS' Soemandjaja won by a margin of 0.5% against his nearest party rival; Central Java III's Gamari and DKI Jakarta III's Adang Daradjatun were two of the candidates with 0.7% and 0.8% victory margins, respectively. Though not so tight compared to other political parties, intense campaigning among PPP candidates occurred in some electoral districts. Among others, Anas Thahir and Zaini Rahman were neck and neck in East Java III, with the latter trailing by just a 0.4% margin. The level of competitiveness among PKB's candidates seems to be higher than PPP. Around 11 of 43 seats received by this moderate Islamic party had a margin of between 0.1% and 3.4% compared to PPP that had less hotly contested battleground constituencies. Among others, PKB's Siti Masrifah won Banten III constituency, defeating her party rivals, including a well-known actor Tommy Kurniawan. The same is also true for candidates running with the Islam modernist party PAN. Almost one-quarter of its total 47 seats in 2014 were decided by a margin of less than 7%.

Overall, the average difference between the vote share of the lowest winner and the losing runner-up's votes from the same party in a given electoral district was 31,801 votes. Table 7.3 shows that the absolute number of votes in the margin of victory in each political party varied slightly from 22,125 votes (Democratic Party) to 39,263 (PAN). In general, these absolute margins of victory in each party are relatively small if we divide by the total number of valid votes cast for all political parties in all electoral districts (77 constituencies). Column 3 of Table 7.3 shows that the average number of valid votes polled in each district was 1,584,463.9 votes.

The results are largely self-explanatory: all political parties suffer from a high degree of competitive intraparty contests measured by a small margin of between 1.40% and 2.48%. The pattern of intraparty competition among Democrats' candidates was highest, perhaps due to the decreasing popularity of the party, which forced its candidates to rely on their personal reputations, as discussed earlier. Meanwhile, though still competitive by any standard, candidates running with PAN witnessed victory with the highest margin compared to other parties. Overall, however, candidatelevel competition in seeking personal votes in Indonesia is comparatively high since it only needs a margin of 1.65% of votes cast for that party on


**Table 7.3** Average margins of victory by political party

Source: Assorted the Electoral Commission (KPU) documents relating to the results of the 2014 legislative elections

average for a candidate to win the final seat won on their party list. This finding is compatible with previous works (e.g. Christensen and Colvin, 2007), suggesting that the level of between-candidate competition in multi-seat districts is likely to be more competitive since the vote share of the lowest winner and the losing runner-up will be much closer to each other compared to elections in single-seat districts.8

#### *7.3.3 Electoral Competitiveness and Vote Buying*

Having discussed the zero-sum nature of intraparty competition, we now turn to examine the relationship between variations in electoral competitiveness and vote buying. Accordingly, I need reliable data on the level of vote-buying incidents at the electoral district level. Pre-election surveys conducted by my polling organisation Indikator before the national legislative election in 2014 are a good source of data for that purpose. In these surveys, multistage random sampling was used to produce a sample that enables us to make inferences and generalisations about the target population. I use 13 surveys with a total number of respondents of 9344, with the numbers per electoral district varying considerably from 410 to 2387 respondents.

Table 7.4 illustrates vote-buying incidence and the average margin of victory in 13 electoral districts.9 The wording used to measure vote buying in these surveys was: "During the run-up to the April 9th 2014 election, did you observe candidates or success team members offering you money, food, household items, and/or other goods (excluding propaganda hats, shirts, and posters)?" Unfortunately, the district surveys were conducted about two months prior to the elections, while vote buying typically takes place or accelerates during the last few days leading up to the polls. We can therefore assume that such practices aren't fully captured by these surveys.

8Again, regarding the electoral effect of vote buying that stood at 10.2% of the electorate, what candidates care most about is how to win the election by a margin that fell within 10.2% of their own personal vote. Recall that it isn't the total party vote that really counts, but the number of voters a candidate distributed money to.

9My polling institute Indikator, along with SMRC, together actually conducted electoral district surveys in 73 out of 77 constituencies. Unfortunately, the wording used wasn't the most explicit version possible to uncover vote buying behaviour. The question was only intended to measure how acceptable vote buying was according to respondents, which of course doesn't allow us to measure whether voters actually accepted electoral bribes or even received vote buying offers.


**Table 7.4** Vote buying and winning margins in 13 electoral districts (%)

But though the available data likely seriously underestimate levels of vote buying, they do allow comparison across a number of electoral districts and therefore help us examine whether a high degree of competition actually drives candidates to engage in vote buying. If this is true, it should be reflected not only in a few days before an election but also some months before the voting day.

To test the impact of electoral closeness on vote buying, I examine the relationship between individual-candidates-level competition and vote buying as a better measure for determining the level of competitiveness in settings like Indonesia where under open-list electoral system, every individual candidate fights for personal votes. Electoral competitiveness in each district should be put in the context of candidate competition *within* parties and *between* parties. This is to gauge the extent to which the level of intra- and interparty competition *within* districts shapes vote buying. Given each electoral district had multiple seats being contested, I include the average margin of all winning candidates across political parties.

In Fig. 7.2, each dot is one candidate gaining the final seat won by his/ her political party. In 2014, of the races for which I have relevant survey data available, there were 92 seats available in 13 electoral districts. At the top of centre point is NasDem's Hasan Aminuddin of East Java II, who won by a large margin in a district whose respondents reported receiving attempts at vote buying at a roughly average rate. At top right is West Java

**Fig. 7.2** Correlations (Pearson's *r*) between competitiveness and vote buying within electoral districts (%)


VII, home of PPP's Wardatul Asriah who won by a fairly high margin in the constituency, where respondents reported the highest rate of vote buying. At bottom left is PPP's Muhammad Iqbal of West Sumatra II, who won by a slim margin in a district where its residents were less likely to be exposed to vote buying.

The dashed line in Fig. 7.2 demonstrates the overall trend. The line falls to the right, again showing that the correlation is in the expected direction, implying that the relationship between competitiveness of elections—marked by smaller winning margins—and vote buying really exists. The Pearson's correlation test returned a significance level of 0.036, proving that the two variables do have a statistically significant relationship. We can be reasonably sure that, as electoral contests grow more competitive, average levels of vote buying increase. It is worth noticing, though, notwithstanding a clear correlation between competitiveness and vote buying, there is a causal issue. This study is unable to assess whether more competitive electoral districts make candidates to buy votes or whether the opposite is true: more vote buying generates more competitive elections.

To sum up, regardless of such chicken and egg problem, the analysis confirms much of the existing literature that states electoral systems shape politicians' strategies and behaviour. When the election of candidates within party lists is dependent upon securing a personal vote, they will respond to such competition by building personal appeals rather than relying on party reputation (Chang, 2005; Carey and Shugart, 1995). Under these circumstances, what matters most in the open-list campaigns is intraparty competition rather than interparty competition. As I have argued, it is the competition between candidates within a party list that makes them engage in more intense personal campaigning (Selb and Lutz, 2015). Given that seats are taken by candidates who obtain the most votes from each list, intraparty competition under the open-list system increases candidates' electoral uncertainty. The dominant narrative among candidates is that they were all dubious about their chances of getting elected—not only rank-and-file candidates placed low on their party list but also party leaders who were placed high.

The link to vote buying is therefore doubly clear: first, when candidates are forced to compete against co-partisans, they can no longer rely on their party label to take them into parliament, and they have clear incentives to differentiate themselves in other ways (Aspinall and Sukmajati, 2016: 13); second, when elections become highly contested, and a relatively small proportion of the overall votes cast can make a difference, strategic investment in vote buying can be expected to alter the outcome of the election. Because the open ballot system only requires candidates to provide a small slice of the vote to beat their co-partisans, the value of each vote increases. Therefore, while vote buying gains a seemingly small percentage of the overall vote, this can be more than enough to help a candidate win in a narrow race. My finding clearly shows that vote buying is an integral part of highly competitive elections. Given the significance of vote buying in determining the final results, my finding slightly differs from previous work by Aspinall and his collaborators (2015b), which suggests that cash handouts in Indonesia are more about meeting an 'entryticket' expectation and less about actual vote choice or turnout. I would argue that it seems to play a deceptively small, but in fact very consequential role in determining electoral outcomes.

### 7.4 Two Alternative Theories

I have endeavoured to answer one of important puzzles underlying this book that revolves around the question of why candidates engage in vote buying while these monetary incentives appear to influence 'only' a limited number of people. As noted above, in closely contested elections such as those in Indonesia, even if vote buying is proven to produce very low returns (due to the problem of voter compliance and broker predation), such a strategy can make a real difference to election outcomes. This explanation, however, leaves a question unanswered: if it were, then we would expect candidates in very close races to spend much more on vote buying, and candidates who expect to win comfortably, to spend less. This leads me to propose two alternative explanations for why candidates might still prefer vote buying, even when the returns are so poor, and even if they aren't in a particularly close race: (1) candidates are trapped in prisoner's dilemma types of situation and (2) vote buying may be an *objectively* inefficient strategy for mobilising votes, but still be *relatively* more efficient than all the viable alternatives.

#### *7.4.1 The Prisoner's Dilemma*

As previously discussed in Chap. 4, as the election drew nearer, the magnitude of vote-buying efforts increased significantly. In January 2014, only 4.3% of voters were exposed to vote buying. As the election period was approaching, however, those who reported being targeted with benefits experienced a sixfold increase with about 25% of voters reporting such exchanges. Not only did the percentage of those being targeted by vote buying rise dramatically but also the incidence of multiple payments increased. Using extensive survey data, I found a consistent pattern in which, as the election drew nearer, it became more likely that a voter had

**Fig. 7.3** How massive were multiple payments in Indonesia? (% of those saying they had been targeted for vote buying). Source: The January, February–March, and late March 2014 data were taken from my pre-election surveys, while the April 2014 numbers were drawn from my post-election survey (see Appendix A)

received payments from more than one party. During the run-up to the legislative elections, I held monthly nationwide surveys asking the respondents, using exactly the same wording, whether they had been targeted for vote buying. A follow-up question asked those who responded affirmatively: "Which party or candidate or success team from which party offered you those goods/gifts?" (Fig. 7.3)

Given that respondents were allowed to give multiple answers, I then recoded the responses to this question to show vote buying by multiple parties. Recall that due to the nature of question and options provided, this exercise can only capture interparty duplication of vote buying efforts, not whether multiple candidates within the same party were providing payments, which was also quite possible due to the open-list PR system. Apparently, candidates believed that a final push could make the difference between winning and losing (candidate from PDI-P, informal communication, 20 April 2014).

The proximity of the election boosts vote-buying incidence largely because of a popular belief among candidates that voters will often vote for the candidate who gave the payment the last and because that candidate will be fresh in the memory. The prisoner's dilemma offers a potential answer for the breadth of this last-minute panic. Candidates might stand to gain more financially if all of them didn't engage in vote buying. But the risk of being trumped for an individual who doesn't participate when others do so, means that such people might feel they have little choice. Applying the prisoner's dilemma to candidates' behaviours when the elections draw near, a candidate will be keen to pursue vote buying if other candidates are using the same strategy. They often see distributing gifts as their best chance of stopping other candidates from winning votes (Guardado and Wantchekon, 2014). For example, a successful candidate from Indonesia's biggest Islamic party admitted to pouring money into the electorate on voting day up until 9 am, having seen an opponent distributing cash just before the polling stations were opened. He defended his actions:

It isn't only a 'dawn attack' [a universally recognised term that reflects the fact that payments are sometimes distributed just after the dawn prayer]. It is also a 'serangan duha' (*dhuha* attack) [referring to *Dhuha* or mid-morning prayer time, which is performed immediately after sunrise when the sun has risen to a certain height]. (Interview, 20 April 2014)

The prisoner's dilemma can be best explained in a context of high competitiveness where uncertainty is great regarding the electoral outcome. Under open ballot systems dominated by personal votes, the personal incentive to win is high, leading candidates to use all available means in their campaigns, including making payments to voters, especially if they see multiple candidates doing it. In the prisoner's dilemma, Takeuchi (2013: 78) points out, "each player's rational strategy to maximise his or her individual payoff ends up with a worse outcome than some other possible outcome that may be better for both players." In the face of closely contested races, the probability of each candidate engaging in vote buying increases, given that small changes in support can be expected to alter the outcome of the election.

In the context of such competitive elections, candidates' strategies are often determined by their competitors' actions. They claim that participating in buying votes, including providing multiple payments to voters, is the best way to tie voters so they didn't turn to cashed-up rivals. As discussed in Chap. 4, candidates and brokers who do this are trying to 'secure' their vote from 'dawn attacks' carried out by other teams. The results from my broker survey confirmed that, when asked to observe other success teams, 28% of the respondents claimed a significant number of other teams delivered cash more than once. As discussed in Chap. 4, many surveyed brokers used the verbs 'to tie' (*mengikat*) or 'to secure' (*mengamankan*) when describing the function of the second payments*,* which is consistent with my qualitative and in-depth interviews with highlevel politicians. Brokers often told candidates to make follow-up cash payments in response to late manoeuvres by rivals. The sense of lastminute panic is evident where the majority of brokers heard that during the cash envelope distribution phase, other success teams were doing the same thing in the same village, but with larger sums of money (see Fig. 7.4). This exactly mirrors the prisoner's dilemma, as discussed above.

Candidates might end up playing the prisoner's dilemma against each other *not only* because of the presence multiple gift givers but also as a response of voters' expectations of monetary rewards. Such voters were accused of being merely driven by money and choosing to vote for rival candidates who paid more. Indeed, there is some anecdotal evidence that voters simply auction their votes off to the highest bidder. One excellent study by Aspinall and Sukmajati (2016) asserts that voters were increasingly ruled by a pragmatic and transactional logic and political actors tried to meet such expectations. Ahmad Muzani of Gerindra told me that a vast majority of voters in his electoral constituency in Lampung I were basically

**Fig. 7.4** Other teams also distribute cash envelopes (%). Source: My survey of brokers, 30 September–25 October 2014

materialistic voters, and that there were numerous reports of locals receiving five to eight envelopes from several brokers. "What can I say? It's the time for them to harvest money (*panen uang*) during election season" (Interview, 13 April 2014). Tubagus Ace Hasan of Golkar reported another colourful story. In his constituency in Banten I, a night before the voting day, the consumption of catfish increased because locals usually hold dinner parties with catfish recipes while waiting for cash envelopes from brokers (Interview, 14 April 2014).

While acknowledging vote buying is a highly uncertain business, and its effectiveness in driving votes is relatively weak, candidates felt insecure about their electoral prospects if they didn't bid on such purely materialistic voters in the midst of rampant handout distribution by other rival candidates. Despite its feeble effectiveness in ensuring victory based on empirical data analysis outlined above, not engaging vote-buying strategies are too great a risk to contemplate as it is widely assumed to be a path to electoral collapse. In this regard, the rationale behind such strategies aren't to 'buy' votes per se, but instead it is widely employed as a countermeasure to neutralise or minimise the electoral gains by their rivals in the context of a highly competitive zero-sum electoral competition. That's why candidates insist on pursuing vote buying, despite such strategy having not always produced the vote that was hoped for, and despite its vulnerability to the problem of broker predation and voter compliance.

#### *7.4.2 Relatively More Efficient Strategy*

The second plausible interpretation on why candidates still choose to buy votes is that this electoral strategy may be fairly inefficient, yielding low returns on investment, although it is still relatively more efficient than all the other feasible alternatives. The fact that Indonesia's candidate-centred election system provides strong incentives for using personal campaigns doesn't automatically imply that money politics will be the only or favoured tactic for pursuing those strategies (Hicken, 2007: 53). In an attempt to win office, candidates might use a variety of means and methods to generate personal votes, including club goods,10 vote coercion, or using media advertising to increase individual popularity. Another personal strategy includes those candidates who have access to state resources in targeting their constituencies. Note that these strategies can interact or occur together at some point in time when implemented in the field (Hicken,

<sup>10</sup>Aspinall and Sukmajati (2016: 23) define club goods as "patronage that is provided for the collective benefit of bounded social groups rather than for individuals."

2007). In Indonesia, for instance, club good provision tends to be combined with other forms of patronage politics, especially vote buying.

I argue that compared with all the viable alternatives, vote buying still emerges as being relatively efficient in driving votes, despite weak monitoring and limited direct yields. Much of the literature on club goods in Indonesia, for instance, shows that such collective patronage, typically consisting of small-scale infrastructure projects or donations to certain associations, is a less reliable strategy in winning votes (Aspinall and Sukmajati, 2016: 23). This corresponds with the popular narrative among candidates that spending money on club goods is a wasteful strategy since they have no guarantees that neighbourhood or community associations that received the benefits will repay the favours with their vote. The provision of club goods is prone to the individual behaviour and perception risk akin to general phenomenon found in the utilisation of common goods called 'tragedy of the commons.' This is basically a situation in a shared resource distributive system where individual voters who receive some benefits as a member of a collective have a perverse tendency to perceptively view the club goods as common collective goods, hence their individual votes tend to be directed elsewhere as they wish or even allowing them not to vote at all since the burden of voting has been shifted to the common or collective efforts. A candidate from PAN, also shared this view, arguing that the provision of club goods was wasteful, in part, because it was ineffective in swaying voters' choices. In his own words:

Many incumbent candidates, long before the elections, had in fact distributed a lot of social assistance to voters, renovated mosques, paid for road repairs, and so on and so forth. But, ironically, they lost to candidates who launched 'dawn attacks' by simply distributing small payments to voters just before the election. Caring for the needs of community is of course important, but it wouldn't be enough. They (the voters) expect to receive concrete, immediate payments. (Interview, 22 April 2014)

Similarly, voter intimidation isn't also a preferred strategy for most candidates in Indonesia because it is simply too risky and costly. That is why, in 2014 election, only a very few people experienced any intimidation from a particular candidate's supporters or campaign team members. My post-election survey of voters found that only 1.7% of respondents admitted being a victim of such coercion. From the perspective of candidates, voter intimidation faces restrictive constraints in terms of money, human resources, and networks (Hicken, 2007). This strategy also has sketchy records in ensuring comfortable victory margin in an electoral contest. Moreover, the use of threat of violence generally carries greater risks compared to relatively safer vote-buying strategy. As already explained in Chap. 2, vote buying has become a common feature in Indonesia's electoral politics, partly due to the problem of law enforcement mechanisms to proceed with such electoral fraud.

Candidates may prefer vote buying instead of or in addition to the use of mass media or social media in supporting individual campaigns. As previously discussed, in the context of candidate-centred elections, the individual popularity is a key determinant electoral success for candidates. Therefore, candidates are expected to use the media campaign to reach out to a vast majority of voters, without completely ignoring the importance of conventional means of campaigning such as direct contact with voters, outdoor campaigns, and other personal strategies such as vote buying. Utilising from a large dataset of voters' surveys in 73 out of 77 electoral districts across Indonesia, I find that the use of mass media in supporting individual campaigns was apparent, but the scale was much less than many observers have expected (i.e. Ufen, 2006), as I will demonstrate in the following pages.

In order to provide a sense of comparison between various methods used especially by incumbent candidates to win personal votes, and to explain the significance of vote buying for candidates to win the seat, I first conduct cross-sectional analysis, that is, comparison among incumbent candidates according to their electoral outcomes (i.e. between the successful and losing incumbent types). In doing so, we are allowed to assess what sort of personal appeals did candidates choose to make in 2014 and voters had most often seen in their neighbourhood? My next strategy is to make a comparison of victorious versus losing incumbents based on individuals' responses towards a variety of means and methods employed by candidates. In my massive surveys in 73 out of 77 electoral districts across Indonesia during the run-up to the 2014 legislative elections, respondents were asked to name which of the following candidates in their respective electoral districts they had most often seen in direct campaign, seen on TV, read in the newspapers, heard on the radio, seen or read on the Internet/twitter/SMS/Facebook called social media campaign, or seen on the banners or posters called outdoor campaign.11 In addition,

11This can be done only for incumbent candidates who possess all variables being tested. Accordingly, there were 339 incumbent candidates who can be analysed for purpose of this study, accounting for about 68% of 502 incumbents seeking for another term in office in 2014. Admittedly, this study lacks representation from the incumbents competing in Eastern part of Indonesia.

these massive surveys measure vote-buying *acceptability* by asking the respondents: "As an effort to win the legislative election, certain candidates or campaign team members gave money or gifts for people to influence their votes. In your opinion, does the money/gift can be considered as something acceptable or unacceptable?" While the wording doesn't allow us to measure whether incumbents actually distributed cash handouts and voters subsequently accepted, it does allow us to measure how prevalent voters' expectation of vote buying is in one's electoral district.

Table 7.5 displays mean scores of individuals' responses about what sort of personal strategies that winning and losing incumbents made and had been most observed by voters. In general, successful incumbents evidently reached out to more voters through various strategies than the losers. Their campaign materials had greater visibility (i.e. through direct contact, mass media, and social media exposure) than that of the losers, suggesting that they were clearly losing the ground-war campaign. All of this suggests that relative to the victorious incumbents, the losers were evidently having difficulties in reaching voters, a fact which may help explain why they were unsuccessful. Overall, the use of media advertising to cultivate personal appeals wasn't a popular strategy for candidates. The pattern prevails, however, in which the victorious incumbents were likely to gain more exposure than the losers. They even gained more exposure in social media, although the Internet use was still limited in Indonesia.

Overall, however, my study found that most incumbent candidates still resort primarily to the traditional means of campaigning such as direct and outdoor campaigning (i.e. publicly visible posters and banners). Table 7.5


**Table 7.5** Descriptive statistics (mean scores) of a variety of methods employed by winning and losing incumbent candidates in 2014 (%)

demonstrates that people were more likely to be exposed by outdoor campaigns than other means of campaigning. There were two reasons behind the incumbents' preference to traditional methods of campaigning than the use of media campaigns. First, this study found that the regular access to the print media, radio, and social media is still limited in most electoral districts in Indonesia. Out of the already limited readership figures for newspapers and small numbers of people who tuned in to the radio broadcast, adding to it those who have Internet access, most of them were not drawn into political news in the first place. This study suggests that investing large sums of money in expensive media campaigns may be wasteful for incumbents. Second, placing advertisements in television is still more beneficial as it has extensive coverage area of up to 90% of the population in Indonesia, although the cost is very expensive. Moreover, television exposure isn't suitable for candidates running in a particular electoral district because they don't need to penetrate wider audiences. Therefore, it is an unsurprising fact when the majority of candidates don't rely on the media as a major means of campaigning, but instead relying simply on placing campaign banners and posters as well as face-to-face meetings.

Interestingly, we see a marked difference too in mean scores of individuals who professed that vote buying is acceptable between the winning and losing incumbents' constituencies. The winning incumbents were more likely to run in areas where voters are actually more accepting of vote buying than the losers. The finding provides suggestive evidence that successful incumbents were found to run disproportionately in places where voters perceive vote buying as a normal occurrence during elections. The analysis of variance (ANOVA) analysis further confirms the finding that successful and unsuccessful incumbents generally have significant different characteristics in terms of the level of vote-buying acceptability in their respective districts. Using a logistic regression model (see Appendix F), vote buying—defined as incumbents' electoral districts whose voters have no problems accepting cash or a gift—stands out as a key determinant of electoral success for their re-election bid. Its substantive effect reaches statistical significance at the 95% level, and it is independent from the influence of other factors included in the equation (i.e. gender, position on party list, name recognition, running in the same electoral district, and district magnitude). A one-unit increase in the level of vote-buying acceptability in the incumbents' districts resulted in an increase of about 2–3 percentage estimate point in the likelihood of being re-elected.

The results from these massive electoral district surveys of voters are consistent with the candidate narrative. In my winning candidate survey, almost 70% of respondents opined that the voters in their constituencies were increasingly pragmatic at election time, expecting monetary incentives in return for their votes. As argued by Hicken (2007), such cultural norms of gift giving not only create a social atmosphere conducive to vote buying but also make this strategy a more appealing avenue for candidates relative to other personal strategies. Although there is accumulating evidence that electoral handouts are oftentimes not quite an effective strategy due to the mistargeting story and brokers' rent-seeking behaviour (see Chap. 5), but nevertheless, relative to other personal strategies, vote buying remains to be the more effective one in securing votes and actually help politicians win.

#### 7.5 Conclusion

This chapter has endeavoured to show the effect of vote buying on voting behaviour in Indonesia. It began with the puzzle about misdirected targeting of vote buying and brokers' rent extraction that might undermine its impact on vote choice. If such a strategy is largely ineffective, why would candidates invest scarce resources in it? But, if vote buying is truly effective, how big or small is the effect and what does it mean and for whom? I have shown in this chapter that vote buying is indeed effective in producing both greater turnout and greater vote share, but the effect is limited to a small minority of voters. In terms of the effect of vote buying on turnout, I have shown that exposure to clientelism has a positive effect on the likelihood of turning out to vote. Rates of electoral participation are significantly higher among those who received cash handouts than those among who didn't (81% vs. 74%).

Regarding the effect of vote buying on voting choice, I have demonstrated that the estimated effect of such practice lies at 10.2%. Note that the 10.2% effect of vote buying is estimated based on those who experienced vote buying compared by the total electorate. If we specifically focus on the effect of vote buying *among* those who received payments from candidates (25% based on a direct survey item), its impact on vote choice was up to 41.8% of the recipients. Hence, in fact, the percentage of the total number of those whose votes can be bought must have been higher if machines are capable of handing out money to voters to more than a quarter of the whole electorate. But, in total numbers, the electoral effect of vote buying in legislative elections was 'only' 10.2% of the whole electorate.

My results answer a critically important question: if it is true that vote buying has a relatively trivial effect—in the sense that it only affects the voter choice of about 10.2% of voters—why do politicians insist on pursuing such a strategy? I showed in Chap. 2 that vote buying has become an increasingly prominent electoral strategy. But judging the *effect* of such vote buying without contextualising it within the context of the electoral system where candidates compete and interact can be difficult. What we really want to know isn't 'how significant is the effect of vote buying,' but 'is it significant enough to achieve a desired outcome?'

Therefore, I have demonstrated at length in this chapter that the seemingly trivial effect of vote buying on voting choice in fact is quite large enough to frequently determine the outcome of electoral races in Indonesia. In an environment where elections are shaped by intraparty competition like Indonesia, candidates depend on personal votes to defeat co-partisans. Under such circumstances, electoral uncertainty regarding the electoral outcomes increases. To measure this, I use victory margins to assess how competitive parliamentary elections in 2014 were in each political party. The empirical evidence reviewed in this chapter reveals that the average margin of victory within political parties—by which winning candidates defeated their internal party rivals—was only 1.65%. Here lies the key to why vote buying remains an attractive investment and has been widely practised in Indonesia. While the effect of vote buying on voter turnout and vote choice may appear small, in Indonesia's highly competitive election settings, that 10.2% matters significantly. The marginal value of each voter collected through buying votes is high enough to constitute narrow winning margins, which helps explain why candidates pursue vote buying, despite its seemingly small effects on voting behaviour.

However, the margin argument raises a further question: why there are still many competitive polities where the vote margins are small, candidates don't engage in vote buying? I have argued that candidates might still prefer vote buying, even when such practice is proven to produce low returns, in the midst of massive money politics distributed by other rival candidates. In a context where multiple candidates were engaged in vote buying, anxious candidates poured money into handout-responsive voters. As such, candidates' decisions to intensify vote-buying efforts were often like last-minute panic buying. What matters isn't whether candidates actually buy a vote, rather that vote buying is a counter-instrument to neutralise their opponents' strategies. In addition to being trapped in a prisoner's dilemma, candidates insist to buy votes because this strategy although proved ineffective in yielding significant votes—but, nevertheless, relative to other personal strategies, vote buying still produces votes that may be more than enough to secure a victory.

#### Bibliography

Books, Journal Articles, and Unpublished Papers


Money Politics, Patronage and Clientelism at the Grassroots. Singapore: NUS Press.


Articles and Reports

Abstract Politics. (2008) Review of "Vote Buying or Turnout Buying? Machine Politics and the Secret Ballot" by Simeon Nichter. Viewed at http://abstractpolitics.com/2008/05/vote-buying-or-turnout-buying/. Accessed 3 April 2017.

**Open Access** This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

# Conclusion

With the rise of democratic regimes in many parts of the world over recent decades, scholarly attention has increasingly turned to assessing the quality of democracy (Kramon, 2013: 252). One of the important measures of the quality of democracy is free and competitive elections. A widely held view among scholars is that electoral clientelism has become a major impediment to such democratic elections. Indonesia is a perfect example. It has made important advances towards democratic consolidation with four consecutive national elections since 1999 and thousands of local elections since 2005. In reality, however, the development of democracy in Indonesia has been burdened by pervasive forms of patronage distribution, especially vote buying. This is especially the case since the introduction of the open-list proportional voting system that has opened a Pandora box of massive money politics within the democratic system. Consequently, the ill-suited design of electoral system has aggravated the accountability of democratic institutions and policy representation in Indonesia.

As a former student activist who took to the streets along with thousands of protesters demanding democratic reforms in 1998—which ended in the birth of a newly democratic regime—in making this study I have been in part motivated by normative concerns about the impact of vote buying on the quality of democracy in my country. More importantly, this study has also been driven by the strong impression I received when I was conducting 13 months' fieldwork in Indonesia. I increasingly felt that the dominant literature on vote buying and turnout buying, which were developed in other settings, was simply insufficient to explain the ubiquity of vote buying in Indonesia. Much of the scholarship on vote buying is based on the Latin American experience. As a consequence, theories of vote buying assume that *parties* determine how money is distributed. In Indonesia, by contrast, *individual candidates* do the vote buying. In Indonesia's open-list ballot system, intraparty competition is fierce, and candidates see party peers as their most proximate threat. Under these conditions, candidates are compelled to rely heavily on personal networks rather than party machines. Intense competition between a large number of candidates also means they only need a small share of the popular vote in order to win office. This differs from the situation in Latin America, where candidates and parties tend to pursue majorities or large constituencies. For all of these reasons, the established theories can offer only limited insight into the machinations of vote buying in Indonesia.

Under such different conditions, and in a context where candidates in Indonesia are largely insecure about their prospects of victory, I have argued in this book that candidates have clear incentives to use vote buying as a means of chasing a small margin of victory. In settings where elections are shaped by intraparty competition as a result of the open-list system, candidates busy themselves fighting on the ground against their internal party rivals in the hunt for personal votes. The desire to defeat their co-partisans makes them risk averse when selecting targets for their material rewards. As a result, most politicians and brokers say that their preferred targets are partisan, loyalist voters, which is strongly evocative of the core-voter strategy that has been devised in the Latin American context.

Yet, as I explained in Chap. 4, my voter survey showed that while such partisan voters are more likely to be targeted than non-partisan voters, in reality the vast majority of vote buying—in absolute terms—happens among undecided voters. This is particularly the case because the aggregate level of mass partisanship in Indonesia is relatively small. Only 15% of Indonesians feel close to a party. This limited number of partisans is also highly contested among internal rivals. Accordingly, my findings show that although politicians tending to target constituents who they think are truly loyal, most brokers end up distributing to voters who receive benefits but don't reciprocate with votes. For various reasons, therefore, we cannot conclude that Indonesian electoral clientelism is based on the pursuit of party loyalist strategies.

In order to explain this combination of features, in Chap. 5, I offered an additional account to the scholarly literature on vote buying by combining the party loyalist model with a role for personal networks. I argued that in Indonesia, candidates and brokers actually intend to target partisan voters, but in reality they mostly distribute patronage to people who are connected to personal networks. Though they think of these people as 'loyalists,' in fact, they might lack any sense of loyalty to the candidate. I call this mixture a 'personal loyalist' approach. Though it acknowledges that candidates largely depend on personal networks to identify voters to target with vote buying, this approach doesn't rule out the importance of party loyalists, seeing also a significant role for personalised partisan voters (i.e. those who possess a sense of loyalty to both the party and individual candidate within the party).

The personal loyalist approach is best explained in the context of openlist systems, as indicated earlier. Under such circumstances, the pressure to collect personal votes is intense. Candidates seek to personalise party constituents to defeat their internal party rivals. However, given only a tiny proportion of Indonesians are aligned with parties, such intraparty competition pushes candidates not to depend solely on party loyalists to win elections. To be sure, despite their limited numbers, such voters are also fought over among multiple candidates from the same party, but their numbers are limited. Accordingly, having personalised their party constituents, politicians seek to expand their electoral base and extend their votebuying reach through personal connections mediated by non-party brokers. In the process, candidates frequently confuse personal loyalists as partisans, and misconstrue people with personal connections to their brokers as loyalists, too. At the same time, expanding electoral bases in this way can make vote buying susceptible to broker predation. Agency loss produces massive rent-seeking behaviours by brokers, making the problems of vote-buying distribution severe (see Chap. 6). It is often the case that brokers exaggerate the number of loyalists, even deceive their candidates on this issue, so that they can engage in predation. As a result, many of the people who are identified through personal networks mediated by brokers are in fact not even loyal to the candidate. These two factors in combination—confusion of personal connections with loyalty and agency loss—contribute to the large amount of targeting of uncommitted voters revealed by my study.

Interestingly, notwithstanding these factors, candidates still insist on spending a large amount of money on vote buying to pursue such spuriously loyalist voters. Recall that as many as a third of voters nationwide are exposed to the practice, making the aggregate level of vote buying in Indonesia the third highest in the world (Chap. 2), as indicated by voter surveys taken over the last decade. Yet these problems, plus other factors, clearly undermine the effect of vote buying on electoral behaviour. Offers of money 'only' influenced the vote choice of roughly 10% of the total electorate (see Chap. 7). Here lies the key to why vote buying remains so important: while this effect may appear small, in Indonesia's highly competitive electoral landscape, that 10% matters immensely. Across the country, the average margin of victory by which winning candidates defeated their co-partisans was only 1.65% (with the winning margin here defined as the percentage of votes cast for a party in a constituency which separated the lowestplaced winner from the highest-placed loser on a party list). The 10% swayed by cash are more than enough to make a difference to electoral outcomes, both in the aggregate and in the case of many individual races. Most politicians, therefore, feel vote buying can play a decisive role in determining the electoral outcome that counts: whether or not they win a seat in the legislature. As a result, many pursued this strategy with enthusiasm.

Chasing a narrow margin victory, however, isn't the only possible answer for why candidates still engage in vote buying, despite high levels of leakage. The prisoner's dilemma offers a potential answer for the widespread practice of vote buying in Indonesia, even if such strategy is proven to produce low returns, and even if they aren't in a close battle. Fearing that their opponents will distribute handouts, candidates might still find it in their interest to engage in vote buying (see Chap. 7). Every candidate might be better if no one bought votes, but if just one person defects and buys votes, then everyone else loses. When the fate of candidates relies on the others' actions, vote buying is often seen as their best chance of stopping other candidates from winning votes, despite its inefficiency of delivery and wastage. This corresponds with the narrative among candidates and brokers that despite vote buying often appears to be *objectively* inefficient, such tactics still be *relatively* more efficient than other electoral strategies (Chap. 7). These three explanations—vote buying as a function of narrow victory margins, politicians end up facing a prisoner's dilemma, and the perceived significance of vote buying in driving votes, relative to other electoral strategies—in combination contribute to explain how and why vote buying is so prevalent in Indonesia.

#### 8.1 Theoretical Implications

I have argued that the personal loyalist strategy helps to explain patterns of vote buying in Indonesia. What are the primary theoretical implications of this personal loyalist approach and how do they contrast with predictions of existing theories? Table 8.1 provides a stylised summary of five major models of vote buying: my own personal loyalist model, plus the following four models: swing-voter model, core-voter model, informational model, and norms of reciprocity model.1

The swing-voter logic predicts that uncommitted voters or weakly opposed voters would be the preferred target of campaign largesse in order to persuade them to vote for the benefactor party or candidate and change the game (Lindbeck and Weibull, 1987; Dixit and Londregan, 1996; Stokes, 2005). This is based on the underlying assumption that a loyal voter is already captive. In contrast, the core-voter hypothesis argues that parties and candidates tend to target their own party supporters because such voters are the most predictable source of votes. The rationales behind the 'core-voter' model vary, ranging from risk aversion on the part of candidates (Cox and McCubbins, 1986), mobilising lukewarm supporters for turnout (Nichter, 2008), the endogeneity of partisan loyalties to electoral handouts (Diaz-Cayeros et al., 2012) to broker predation (Stokes et al., 2013). Unlike the swing- and core-voter models whose key parameter in distributing benefits largely depends on voters' partisan proximity to the machine or to its opponents, the norms of reciprocity model doesn't require voters to have strong ideological attachments. Instead, according to this model, clientelistic practices produce a sense of moral obligation or indebtedness on the part of beneficiaries to vote for the distributing candidate in exchange for the reward (Finan and Schechter, 2012). Finally, drawing from much of the African elections, Kramon (2013) developed the informational model of vote buying, arguing that vote buying is a mechanism for politicians to establish credibility with voters regarding the distribution of patronage and private goods in the future. It serves primarily to convey information to voters that candidates are credible and able to provide future rewards.2

1For comparison, see Kramon (2013: 65–69), who also made some comparisons between his own informational model and other three major approaches to the study of vote buying: the swing-voter model, the core-voter model, and the norms of reciprocity models. In some ways, I also adapt from Kramon's comparisons.

<sup>2</sup> In addition to Kramon (2013), the informational argument can also be found in the works of Lindberg (2003) and Nugent (2007).


B. MUHTADI

Each line of Table 8.1 shows what theoretical preconditions or outcomes are to be expected by each of these models. The first group of 'Contextual Factors' sets out the features of electoral competition under which the model is assumed to be applicable. The second, 'Nature of Vote Buying' summarises what form vote buying takes in each model. Finally, the table summarises effects of vote buying in terms of both vote choice and turnout.

For simplicity's sake, let's focus here only the aspects of my personal loyalist approach that have the most notable theoretical implications. The first distinctive feature of my model concerns the political context, specifically the role of political parties. I have demonstrated from the outset that the existing theories of vote buying have emerged in contexts different from that in Indonesia. Table 8.1 makes these differences visible in a number of fields, such as party organisation, electoral system, party base, partisanship, and voting system. In particular, much of the extant literature on the swing- and core-voter models relies on the underlying assumption that party machines have the capacity to enforce vote-buying agreements, and that they engage in ongoing constituency service (Stokes, 2005; Kramon, 2013). In contrast, the other three models (i.e. the norms of reciprocity and informational models as well as my personal loyalist model) don't require well-organised party organisations. In Kenya, as in many African countries where the informational theory was produced, the main proponent of the approach Eric Kramon (2013) shows that socially embedded and well-organised parties don't exist. Similarly, the reciprocating model is built on the assumption that political parties are weak. Drawing from an excellent study conducted in Paraguay, the key theorists of the model Frederico Finan and Laura Schechter (2012) demonstrate that political parties in the country are weakly organised, not strongly ideologically oriented, and less embedded in society. Accordingly, Paraguayan politics tend to be extremely personalised (Rizova, 2007). My personal loyalist model arises in a similar setting. Vote buying in Indonesia, therefore, doesn't rely upon strong party organisations but can also be mediated through a variety of informal networks. This study has provided evidence that institutional factors matter. The non-party organisation of vote buying in Indonesia is largely a product of the open-list system, which incentivises intraparty competition and prompts candidates to invest in building campaign teams that rely on personalised networks.

Still with regard to the role of parties, much of the literature on swingand core-voter models broadly assumes that *only* one machine has the ability to engage in clientelist exchange (Nichter, 2010). This assumption may be justified in the context of contemporary Latin America, where mostly only one party can take advantage of the state resources and social networks required for clientelism (Stokes, 2009: 12; Nichter, 2010: 97). This, however, doesn't really fit in Indonesia where there is cartelised party system, so that no party is locked out of state resources (Slater, 2004; Ambardi, 2009). Note that all parliamentarians independent of party affiliation have opportunities to access resources allocated for their constituencies (Farhan, 2016). More importantly, the expectation of the personal loyalist strategy is that multiple candidates from the same party or other parties compete against one another even in the same neighbourhoods or households. This runs contrary to the existing literature on vote buying, which assumes that each machine tends to cultivate separate networks with distinct constituencies (Stokes, 2005: 324; Gans-Morse et al., 2014: 17). In Indonesia, by contrast, as Aspinall and his collaborators (2017: 2) put it, "competing network machines often overlap within the same geographical locations and social milieus." Many areas, especially in Java, had become free-for-all battlegrounds in which multiple candidates fight for personal votes in every village, neighbourhood, and laneway (Aspinall et al., 2017: 12). Multiple candidates from multiple parties also compete against each other to recruit brokers and determine their base areas. In sum, borrowing Stokes' (2005: 324) words, these 'duelling machines' not only compete to purchase votes, but they also avidly bid up the price of votes to outbid rivals (see Chap. 6; also, Aspinall et al., 2017).

A second distinctive feature of the personal loyalist model concerns the purpose of the vote-buying transaction. The swing-voter model assumes that vote buying is an exchange of a reward for a vote choice: a voter receives money and in return votes for the giver. Conversely, according to the core-voter school, the payment isn't to 'buy' a vote, but rather to mobilise supporters to turn out. In the norms of reciprocity model, the expectation is that such exchange serves either in a model of vote buying (persuading swing voters) *or* turnout buying (mobilising core voters). The informational theory views the transaction as a mechanism to convey a signal of candidate credibility with respect to future performance (Kramon, 2013). The personal loyalist strategy offers a slightly different story. In Indonesia's extremely competitive election settings, where only the winner takes home the prize of office, vote buying serves as a means of providing a small margin of victory. Minor shifts in support whether as a result of buying lukewarm supporters for turnout (core-voter model), or purchasing the support of uncommitted voters (swing-voter model) can make a difference in the electoral outcome. Most politicians in Indonesia realise that vote buying doesn't ensure victory in an election, but they do believe that it increases their chance of winning in a closely contested election. The dominant narrative among politicians is that if they don't engage in vote buying and others do, they will certainly lose. Hence, although vote buying doesn't always produce the vote that was hoped for, candidates still have incentives to pursue vote buying because in the context of fierce campaigns like Indonesia the value of each vote collected through such exchange can potentially make the difference between winning and losing.

A third point where my thesis about the personal loyalist strategy differs from other approaches is with regard to targeting strategies. Given that the primary purpose of vote buying under the swing-voter model is more to sway voters' decisions than increase their turnout, the theory therefore predicts that politicians tend to target voters who are ideologically unattached or weakly opposed supporters. In contrast, the core-voter theory observes that politicians will try to target lukewarm supporters to persuade them to turn out on voting day. While the swing- and core-voter models employ an ideological test to explain who is or isn't targeted by clientelist parties, the reciprocity model predicts that political machines target voters who are intrinsically reciprocal. According to this line of reasoning, machines will not distribute their largesse randomly across the electorate but will look at whether voters feel indebted to reward those who have helped them in the past (Greene and Lawson, 2012). By contrast, the informational model argues that the targeting of handouts will be relatively diffuse. Given that under such a model vote buying is a mechanism to signal credibility with respect to future rewards, cash distribution is therefore less targeted at specific types of voters (Kramon, 2013). In Kenya, where the literature on the informational model is based, distribution of cash and gifts can take place in public and one voter can receive multiple payments from multiple candidates, a strong indication that vote buying is distributed in a diffuse manner. An expectation of my own approach is that politicians and brokers express strong intentions to target constituents they think are truly loyal, but in reality, they mostly target uncommitted voters who will not always reciprocate with support. This is largely because the concept of loyalty in Indonesia is ambiguous and has multiple dimensions ranging from kinship, ethnic and religious ties, receipt of patronage to connection via brokerage networks. When candidates and brokers claim to be targeting partisan, loyalist voters, they don't only rely exclusively on partisan loyalties but also judge the target in terms of personal networks. As a result, they misidentify non-partisans as partisans because they assume personal connections are partisan leanings. But the key is that they target persons who are connected by personal networks to their brokers, even if some of these persons, in fact, lack even a sense of personal loyalty to the candidate.

As with the informational model, my approach contrasts with the assumptions that underlie the two dominant theories of vote buying, which suggest that politicians distribute particularistic rewards in a highly targeted way to specific types of voters guided by the partisan preferences of the recipients to the machine or to its opponents (Diaz-Cayeros et al., 2012). Furthermore, both the swing- and core-voter models require ideological parties and ideological voters (Amick, 2016). In contrast, it is extremely difficult in Indonesia to target individuals based on their political preferences, not only because the population is becoming less attached to parties, with only one in every ten people in the country who feel close to a political party (see Chap. 4) but also because parties are becoming less ideological and are weakly rooted in society (see Chap. 1). Tomsa (2010) claims that most Indonesian parties are now presidentialist in essence and no longer represent sharply defined ideological constituencies. The current models of vote buying were developed in a setting where some parties possess very strong 'base areas'—for example, working class areas in Argentina that have been voting for Peronists for decades, or the long-time rural voter base of the Party of the Institutionalised Revolution (PRI) in Mexico (see Chap. 5). A few parties in Indonesia, like PDI-P and PKB, have quite a strong tradition of having partisan voters; most don't. All of this distinguishes Indonesia from the Latin American cases on which the literature is based and therefore explains why the targeting of vote buying is distinctive in Indonesia.

Finally, existing theories also differ with regard to the impact of vote buying on voting behaviour. Most existing models predict that the exchange will influence vote choice; only the core-voter approach differs in this regard, instead focusing on mobilising the turnout of passive loyalists. Although the dominant theories of vote buying admit the effectiveness of such a practice on vote choice, they have no consensus in the answer to why vote buying sways individual's vote decision (Kramon, 2013: 68). The expectation of the swing-voter model is that despite targeting those who are ideologically unattached, machines are able to detect who has kept with the vote-buying bargain and enforce the deal (Stokes, 2005: 322). As noted above, in this view, parties have the capacity to monitor the recipients and ensure they reciprocate with votes because they are bottom-heavy and socially embedded in local communities (Stokes, 2005). In the reciprocity model, vote buying is effective because it is a moment of retrospective evaluations of candidates, when voters feel a moral obligation to vote only for those who have provided them with rewards (Kramon, 2013; Greene and Lawson, 2012) and because of their hopes of maintaining close social relations and receiving future rewards (Finan and Schechter, 2012). Meanwhile, the informational model has little to do with retrospective evaluations (Kramon, 2013). Instead, the framework largely depends on prospective expectations of politicians, where voters expect to receive patronage goods in the future.

Overall, the dominant models assume that cash handouts will affect vote choice and motivate the recipients to turn out on voting day. Though my study doesn't directly address this debate, it does take up one issue which has been largely neglected by the scholars who have engaged in it. Thus, scholars proposing these approaches have largely neglected the critical issue of the *magnitude* of the effect on vote choice and turnout: how big is the effect of cash gifts on both vote choice and turnout and how should we interpret it? The conventional literature on vote buying generally fails to quantify the effect, assuming that vote buying will automatically result in higher turnout or vote share. The expectation of the personal loyalist strategy is that the impact of vote-buying exchanges may appear insignificant in numerical terms but could still be influential in determining electoral outcomes. Recall that the meaning of the effect of vote buying, whether small, medium, or large, varies by context. A 10% effect on voting decisions might be small in some contexts, but this number will be sufficiently high to clinch a victory by a vote buyer in many highly competitive election settings, as in Indonesia. This is similar to how advertisements work. They might be unable to make all viewers buy the product being advertised, but at least some people get interested, translating into increased sales. Likewise, vote buying may be 'ineffective' in yielding significant votes to the extent anticipated by the buyer, but the votes that are flipped may be more than enough to make him or her win.

### 8.2 Policy Implications

Having discussed some theoretical dimensions of this study, we are now in a position to review the policy implications. This book clearly has implications for democratic accountability, in general, and reform of electoral institutions, in particular. In Chap. 1, we began the research by delving into an explanation of institutional arrangements in the post-Suharto era, seeing these as shaping the supply side of vote buying. Changing electoral rules clearly shapes the propensity of candidates to engage in vote buying. Therefore, it is appropriate for this study also to offer 'supply-side' remedies to discourage parties or candidates from adopting vote-buying strategies.

This study found that patronage distribution has been central to election campaigns in Indonesia. The results demonstrate that most candidates pursue vote buying because they see this strategy as affecting the outcomes of many competitive elections, where small changes in the voting calculus can alter the final results. An obvious conclusion is that the open-list system, in which a small number of personal votes are expensive yet critical for politicians, has been responsible for the growing prominence of cash handouts during political campaigns in Indonesia. As discussed in Chap. 1, vote buying was unheard of in Indonesia's first free election after the fall of Suharto in 1999 when Indonesia adopted a fully closed-list system (where voters can only vote for parties and don't have power over the order in which party candidates are elected). The practice of vote buying began to flourish in the 2009 election, after the introduction of a fully open-list system in that election. The fact that each candidate not only competes with candidates from other parties but also with other candidates from his or her own party has exacerbated the practice.

Although this list isn't exhaustive, we can speculate that the open-list system has three far-reaching implications for Indonesia's political system as a whole. First, the open-list system obviously makes elections more candidate-centric because they create incentives for the cultivation of a personal vote. The results in Chap. 1 show that official statistics of parliamentary elections over the period of 2004–2014 exhibit a clear linear trend towards an increase in the share of personal votes cast and a decline of party votes. The finding from a series of voter surveys conducted by my polling institute, the Indikator, in 56 electoral districts in February 2014 with a total of 43,510 respondents also shows that candidates have become more important. Likewise, a voter survey conducted by LSI and La Trobe University in North Sulawesi and Maluku in October 2012 suggested that voters were more likely to vote based on specific candidate attributes than on the party affiliation of the candidates.3 This corresponds with the results as presented in Chap. 6 that the candidate's qualities and personal reputation are far more important than the programme of the party for which the candidate runs. The increasing role of candidates should be read in the context of the open-list PR system with its emphasis on intraparty competition (see Chap. 7). Under such circumstances, candidates are forced to compete against each other and differentiate themselves from their internal party rivals, including by vote buying. Accordingly, they need to run well-structured but expensive success teams and expend money on cash payments, club goods, and other handouts.

Second, it follows that given that the open-list elections create incentives for candidates to pursue clientelist strategies, money has become the most important foundation of political success. Indeed, more money doesn't guarantee victory, but it does increase the chance of it (Aspinall et al., 2015). This was very evident during the 2014 campaign, which most candidates dubbed as the most 'brutal' election in Indonesian history. Zuhairi Misrawi, a DPR candidate from PDI-P with NU background jokingly put it, "there is a new Islamic jurisprudential maxim in politics: '*Al-fulus tuhyin nufus, ma fi fulus manfus*' (money will extend your life. If you don't have money you will die politically)" (Informal Communication, 2 July 2016). As Indonesia has moved away from partycentred to candidate-centred campaigns, candidates themselves have had to engage in costly mobilisation efforts, including running advertisements, commissioning surveys, building campaigns, mobilising constituencies, and buying votes. The popular view is that given such massive costs, only better-resourced candidates can do well in elections. These expensive electoral processes, coupled with the fact that most parties are less ideological, mean that parties are open to nominating candidates from outside the ranks of their own cadres. The Forum of Citizens Concerned about the Indonesian Legislature (FORMAPPI) revealed that in the 2014 election, only 33% of the candidates could be classified as party cadres. Almost half of the total candidates (3241 out of 6607) running for national parliamentary seats had business backgrounds, with

3The candidate-centred campaigns have been particularly strong in Eastern Indonesia partly because the party systems in the region tend to be less patterned and less institutionalised than those in the western part of the country (Tomsa, 2014: 250).

many joining the parties just a few months before the election (FORMAPPI, 2013).4 The entry of such business people has come at the expense of candidates with backgrounds in political and social activism, who are considered to be unable to fund their own campaigns (Budiman Sudjatmiko, Interview, 29 April 2014). Such findings correspond with those of a study from Centre of Political Studies, University of Indonesia (PUSKAPOL UI) that documented two important aspects of the backgrounds of those elected in 2014. About 58.86% of MPs had backgrounds in business or the private sector, or they were entrepreneurs or professionals,5 suggesting that money is important for winning elections. PUSKAPOL also revealed that 77 out of 560 elected candidates came from wealthy political dynasties (*Republika*, 9 October 2014). Around seven out of them were among the top ten candidates receiving the biggest share of votes in 2014.

Third, by contributing to more candidate-centred elections, the openlist system has also jeopardised the relationship between voters and parties, making party cues less important and directing voters towards the shortterm appeal of candidates. This electoral system has also contributed to the rapid decline of party loyalty (Chap. 5) and diminution of the image of parties in the public eye (Chap. 6). Survey data show that the number of Indonesians who feel close to a political party has declined significantly from 86% in 1999 to a mere 10.1% in February 2019. This number is comparatively low, making Indonesia a country with one of the lowest partisanship levels in the world. Interestingly, the level of party identification started collapsing in 2004 (see Chap. 4), when the country introduced the semi-open proportional system, and for the first time, ballot papers featured candidate names as well as party logos, allowing voters to vote for a particular candidate rather than just for a party. This parallel development of declining party loyalty and the adoption of candidatecentred elections has made it difficult for parties to mobilise voters on the basis of programmatic campaigns and policy positions.

Accordingly, this study recommends changes in institutional design, law enforcement, and voters' education in order to reduce the practice of vote buying:

<sup>4</sup>Forum of Citizens Concerned about the Indonesian Legislature (FORMAPPI), "Anatomi Caleg Pemilu 2014," 3 October 2013.

<sup>5</sup>Republika, "Ini Dia Profil Anggota Legislatif 2014–2019," 9 October 2014.

1. Changes in the electoral system from open-list proportional system to closed-list proportional system. In order to reduce vote buying and move towards more programmatic politics, Indonesia needs to also move towards a more party-based system of electoral competition. This can be made possible by revising the current electoral system and returning to closed-list proportional representation. Shifting the arena of competition from intraparty competition between individual candidates, to interparty contests, should reduce the need to generate personal votes through vote buying. More importantly, in closed-list multimember districts, voters choose among parties, and the rank order of the candidates in the party list is determined by the party. Given that citizens aren't allowed to express a preference for any particular candidate within each list, this type of electoral system tends to produce party-centric elections. In such a system, voters should increasingly turn their attention to party policies rather than personalities, enhancing party cohesion, reducing internal disputes, and centralising party leadership (Suwarso, 2016). As Norris (2006: 105) argues, closed-list elections "encourage politicians to offer programmatic benefits, focused on the collective record and program of their party, and to strengthen cohesive and disciplined parliamentary parties." In an environment where voters have less choice to determine the fate of individual candidates, and where campaigns are more focused on party platforms than on personal reputations and connections, we can expect a reduction in the importance of money in determining electoral outcomes.

Timor-Leste is an interesting example of how the closed proportional system reduces the incentives of parties or candidates to use bribery tactics in elections. Although Timor-Leste's voter-side demand for vote buying is high, where 32.7% of citizens consider such strategies are acceptable in the 2017 legislative elections, the incidence of vote buying in that country was considered as very rare occurrences. Only 4% of voters claimed to be targeted by money politics.6 This happened because the closed proportional system used in Timor-Leste has made political parties the main players during the election. As a result, the supply side of vote buying is limited

<sup>6</sup>The Timor-Leste Survey Institute and LSI, "Exit Poll: Eleisaun Parlamentar Timor-Leste," 22 July 2017.

by design. Of course, the closed proportional system isn't without its flaws. In the 1999 elections in Indonesia, this system was proven to encourage elite oligarchs. Additionally, there were numerous reports of wealthy candidates purchasing winnable slots on party lists by bribing party leaders. But even accounting for this weakness, the diminished incentive for widespread vote buying in a closed-list system makes it preferable to the easily exploitable status quo. The closed proportional system is, at least, able to isolate and confine the incidence of vote buying at the party elite level instead of at the grassroots voter level, increasing the role of the party in electoral battles which ultimately leads to increase in the institutionalisation of political parties. In order to reduce elite oligarchs, a closed proportional system can be accompanied by primary elections or conventions within the party to screen the respective legislative candidates.


remaining seats in higher magnitude electoral districts due to the shrinking district magnitude.


Overall, this book has argued persuasively that electoral systems matter in explaining the ubiquity of vote buying. I have shown that Indonesia's open-list voting system is creating a situation where one-third of the electorate engages in vote buying. Accordingly, in the context of transition stages towards a mature democracy, one has to cautiously attend the party and electoral system. Especially within new democratic countries, the political parties are still in a generally weak and vulnerable state. If there is some incongruous predicament between the party or electoral system and the current condition, it would only result in long-term damage towards the political system, that is, sluggish transition and feeble party institutionalisation, as well as unbridled money politics.

# 8.3 Concluding Remarks

In the last part of this book, I highlight my contribution to the study of vote buying and how it might make a difference. Overall, this study makes four contributions to the comparative literature on vote buying. First, in terms of methodology, this study has used multiple methodologies by combining quantitative and qualitative approaches to determine the patterns of vote buying in Indonesia. Most previous works on clientelist exchanges are either purely ethnographic or they rely heavily on survey data (Kramon, 2013: 253). Focusing solely on one type of approach is susceptible to methodological problems. The general challenge of the former is to establish the extent of clientelism in a population and to determine causal inferences. The methodological problem of the latter is in understanding a phenomenon or the context in which the data are collected. Mixed methods of the sort I used in this study can offset these weaknesses and have allowed me to develop a broader perspective of vote buying in Indonesia. My surveys and qualitative research allow me to determine the scope of vote buying, trace its causal mechanisms, test novel hypotheses, provide interpretations in a meaningful way, and adjudicate between claims and debates in the existing literature. In particular, by relying on voter surveys, my study has demonstrated that the level of vote buying in Indonesia is high by international standards, and such practice is effective in producing slightly—but electorally consequential—higher rates of turnout and vote share that are enough for most candidates to secure victory. No other methodological approach can so systematically establish the extent and the effectiveness of vote buying. In addition, supported by a list-experiment, my study has also showed that direct survey items about vote buying in Indonesia aren't subject to response bias. Moreover, unlike much of the literature on vote buying which largely relies on voter surveys, my analysis isn't based solely on the demand side of vote buying. By utilising broker and politician surveys, my study also puts emphasis on the supply side in order to understand vote buying from the perspective of the actors who orchestrate it.

Second, by embedding my quantitative findings in rich empirical findings drawn from my qualitative fieldwork, I have been able to demonstrate that investigating the political context in which vote buying operates is crucial to really understanding how the mechanism work in practice. Much scholarly theorising on vote buying is based on empirical observations drawn from several Latin American countries. In these countries, as already noted above, the organisation of vote buying is party-based, and political parties aren't only socially embedded, but they are also well organised. The Indonesian case, however, is very different. This study has explained the prevalence of vote buying in the Indonesian setting where partly as a result of the personalised electoral system and intense intraparty competition that occurs there—candidates rely on personal networks rather than their party, where partisan ties are weak, and where personalised loyalties matter far more. I have showed that the meaning of 'partisanship' in the framework of the Indonesian context—where a lot of connections between voters and candidates aren't mediated by political parties but by informal brokerage networks—is far more nuanced than we might expect. One may relate the findings from Indonesia to neighbouring countries such as the Philippines and Thailand, where much of the literature suggests that political parties are generally weak and where votebuying practices are prevalent. Julio Teehankee, for instance, shows that after the Second World War and long period of dictatorship under Ferdinand Marcos, political parties in the Philippines "suffered from weak internal organisation, structure, and discipline, which resulted in weak party loyalties and constant party-swiching" (2012: 190). Schaffer (2007: 3) also shows that in the Philippines, clientelist practices have also become a central feature of electoral politics. However, one major point of divergence is that constituency service in the Philippines plays a greater role in the interplay between parties and voters at the grassroots than in Indonesia (Berenschot, 2015: 560). Additionally, Filipino voters were more likely to be attached to parties with 42% feeling close to political parties in 2005, compared to Indonesian with only 15% of the electorate who did so in 2014 (see Chap. 5). Similarly, despite sharing a lot of similarities with Indonesia with regard to the widespread occurrence of vote buying (Hicken, 2002; Callahan, 2005), weak party organisation, and the use of personal campaign strategies (Hicken, 2002), political parties in Thailand are more likely to engage in constituency service than their counterparts from Indonesia. In her excellent ethnographic research in Thailand, Bjarnegård (2009: 123) shows that although they show little interest in promoting party policy, parties in the country actively arrange drainage, roads, electricity, and admission in school for their party supporters, suggesting that *parties* largely take care of the needs of community which typically happens beyond elections (Berenschot, 2015). In Indonesia, by contrast, *individual candidates* present themselves as caring for the interests of the community and offer 'concrete' benefits to their constituents, especially during election time. This greater role of individual candidates makes vote buying in Indonesia more personal (i.e. non-party-based) and might distinguish it from the existing literature, especially that which stems from Latin American cases.

Third, the fine-grained analysis of this study also contributes a more detailed understanding of how targeting of vote buying in Indonesia works, arguing that empirical evidence reviewed in this project fits differently to the pathway of the lively debate in the literature between proponents of the core- and swing-voter models. My research design, which primarily relies on tracking individual voter surveys, allows us to conclude that although there is a greater likelihood of party loyalists being targeted, the absolute majority of vote buying happens among non-partisans, given the relatively small number of party loyalists in Indonesia. In fact, my rich probability sample of low-level politicians and brokers, and in-depth interviews with national politicians, provide evidence strongly evocative of the core-voter argument. I have highlighted three pieces of empirical evidences explaining the gap between the politicians' intentions to focus on party loyalists and the fact that in total numbers, swing voters are more targeted than loyal supporters. First, candidates and brokers tend to exaggerate the number of partisan voters. Second, loyalty is an amorphous concept and has multiple dimensions in Indonesia. Third, agency loss results in both unreliable brokers and unreliable voters, confirming the classic problem of vote buying as an uncertain business. These are the points for departure for the personal loyalist approach I put forward. Despite candidates and brokers' claims that they were targeting loyal voters, it turned out that they were often providing benefits to basically uncommitted voters.

Lastly, my study has endeavoured to advance our understanding of the logic and motivations behind candidates' insistence on pursuing vote buying, regardless of the fact that the targeting of vote buying is so misdirected, and there are principal-agent problems among politicians and brokers, and between brokers and their voters. I have shown throughout this study that despite such intrinsic problems of clientelist exchange, vote buying remains an attractive investment and has been widely practised in Indonesia. This is largely because candidates believe that the voting decisions of a sufficient proportion of voters can be swayed by rewards. Even if the overall number of such voters is small, it is often more than enough to provide a vote buyer with a narrow winning margin. In addition, applying the prisoner's dilemma to interactions between voters and party machines, the presence of other candidates doing vote buying to win elections may explain why they insist to distribute handouts, even if these actually don't purchase vote. Further, much of the emerging conventional wisdom among candidates is that vote buying is a more reliable strategy in winning votes, relative to other mobilisation tactics. By emphasising the role of vote buying as a mechanism to provide a small margin of victory, and by understanding the electoral motives behind candidates' distribution of rewards, we can better understand how and why vote buying has become so rampant in Indonesia.

#### Bibliography

Books, Journal Articles, and Unpublished Papers


———. (2010) Politics and Poverty: Electoral Clientelism in Latin America. PhD thesis, University of California at Berkeley, USA.


Stokes, S.C. (2005) Perverse Accountability: A Formal Model of Machine Politics with Evidence from Argentina. American Political Science Review 99(3): 315–325.

———. (2009) Pork, by Any Other Name … Building a Conceptual Scheme of Distributive Politics. Paper presented at the American Political Science Association Conference, Toronto, Canada.


**Open Access** This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/ by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

# Appendix A: Indonesian Voter Surveys

As discussed in Chap. 1, the data used in this study are largely based on a national public opinion survey I conducted immediately after Indonesia's legislative election in 2014. Given the centrality of this survey, I explain here its methodology, notably its sample size, sampling scheme, pre-survey preparation, data collecting, quality control, data entry, and data analysis. Note that because I have been involved with the Indonesian Survey Institute (LSI) and Indonesian Political Indicator (Indikator), I, here, simply describe standard survey methods that have been used regularly by my polling institutes; the survey in question followed these standard methods. In addition, this study incorporates trend data from three preelection surveys conducted by my polling institute, Indikator. However, since these pre-election surveys were done by the same survey institute and followed Indikator and LSI's standard sampling methodology and operating procedures, I simply describe their sample sizes, timing of data collection, margins of errors, and numbers of original and substitute respondents.1

<sup>1</sup> In addition to the nationally representative public opinion surveys, this study is also enriched by a series of electoral district surveys and local election polls conducted by Indonesia Survey Institute (LSI), Indikator, and Saiful Mujani Research and Consulting (SMRC), especially when dealing with the issue of vote buying. These three non-partisan research institutes generally use the standard methodology of surveying Indonesians over the age of 17. Importantly, I have explained already the sampling methodology of those surveys, most notably in Chaps. 1 and 2.

© The Author(s) 2019 267 B. Muhtadi, *Vote Buying in Indonesia*, https://doi.org/10.1007/978-981-13-6779-3

# The April 2014 Post-election National Survey

My original survey took place from 22 to 26 April 2014, around two weeks after the 2014 legislative election. This timing provided citizens with a recent reference point for a variety of questions, including whether they had been targeted for clientelist exchanges during the campaign. The population of this survey was all Indonesian citizens who had the right to vote in elections: those who were 17 years old and above or already married when the survey was conducted. The planned sample size was 1220 voting-age adults, but one primary sampling unit (consisting ten respondents) was considered to consist of defective interviews. Overall, the sample size was thus 1210, selected with multistage random sampling, proportionally distributed over the 34 provinces. Based on this sample size, the estimated margin of error is ±2.9% at 95% confidence level, assuming a simple random sampling design. The margin of error is at its highest when the true proportion being estimated is close to 50%. Nonetheless, somewhat higher error margins should be expected because of the use of multistage cluster sampling in this survey. It is worth noting that the design effect isn't easily calculated using established statistical software.

With regard to the sampling scheme, the administrative system in Indonesia is divided into provinces (*provinsi*), districts/cities (*kabupaten/kota*), sub-districts (*kecamatan*), urban/rural villages (*kelurahan/desa*), neighbourhoods (*rukun warga*—RW), and subneighbourhoods (*rukun tetangga*—RT). Since this survey unit was designed to be nationally representative, the demographic composition of the sample gender proportion, province, and rural-urban residence—should reflect the voting-age adults of the Indonesian population based on the 2010 Census (https://sp2010.bps.go.id/). As Indonesia is divided into 34 provinces, the sample was drawn proportionally in each province using random selection of the samples according to the following procedures.

The population was initially grouped based on the population of each province across Indonesia; in this way, we can produce proportional samples in each of the provinces. The second stratification was conducted on the population proportion based on gender. According to the 2010 Census, the sex ratio for Indonesia is 101, which means that for every 100 females, there are 101 males (https://sp2010.bps.go.id/). In this study, the ratio was therefore rounded to 100, and therefore the gender proportion in the sample was equal: 50% male and 50% female. The third stratification is classification which is based on the area of domicile: urban and rural. Following the 2010 Census, the proportion was 49.79% urban and 50.21% rural (https://sp2010.bps.go.id/).

It must be noted, however, that the rural-urban proportion varies by province. Accordingly, the number of rural-urban respondents was selected in proportion to the size of population in each province based on the primary sampling unit, that is, the *desa* (rural villages—the smallest administrative unit) or *kelurahan* (urban villages, wards). Systematic random sampling was done on the villages (urban or rural), selected in each province according to its proportion of population. From each of these primary sampling units, ten respondents were chosen. Our research teams then went to the office of the selected rural or urban villages, and asked for the list of *Rukun Tetangga* (RT, the smallest neighbourhood units or hamlets). All RTs were listed, and then five RTs were selected at random. Our interviewers then met with the RT officials and asked them to provide the list of households (*Kepala Keluarga*, KK) in the selected RTs. Having listed all of the households in each selected RT, two households were then selected at random. In each of the randomly selected families, all of the household members with the right to vote were listed—those who were 17 years and older or married—and one person, male or female, was selected to be a respondent. If the first household member chosen was a female respondent, then the next household member would be male, and vice versa. The interviews were conducted face to face.

With 1220 respondents initially planned, this survey selected 122 primary sampling units (PSUs), in which each PSU consists of ten respondents/interviewees. Hence, one enumerator was assigned to do the interviews in one PSU, so that the numbers of enumerators were 122 persons. In total, I deployed 157 people during the data collecting, including 30 area coordinators and their field assistants and 5 national supervisors. I assigned one field manager to monitor the survey full-time. As I have been working in the Indikator and LSI for several years, I used my institutes' resources, including a pool of experienced enumerators located in every province of Indonesia. Most of them were university students and part-time employees who have been working with my survey organisations for some time on a project basis. They come from various educational backgrounds, such as humanities, social sciences, engineering, and natural sciences. From this pool of enumerators, we selected the interviewers for this survey. Important criteria we usually use for selecting the interviewers are competence, integrity, trustworthiness, independence, and ability to work in a team.

Simultaneously, I developed the questionnaire of this survey. For purposes of generalisation and international comparability, this survey used a number of questions that are partially derived from and comparable to similar questions that have been asked elsewhere. Nonetheless, I had also my own particular questions to adjust to the aspects of the Indonesian situation I was most interested in. Some measures of vote buying employed in this study were generated from the project 'Money Politics in Southeast Asia.' I would like to acknowledge Prof. Edward Aspinall and other principal investigators of this project, who in the spirit of advancing academic research on this topic, generously provided me with some additional funding for conducting this post-election survey and, more importantly, allowed me to employ their measures and share the results for the purposes of this study. The definitive language version of the questionnaire was Indonesian, and if necessary, it was translated into local languages by my interviewers. Given that some questions were taken from a larger project of a four-country study of money politics as noted above, and were written in English, it took a few days to translate them. Pre-testing the questionnaire was undertaken on about 25 adults from different socioeconomic backgrounds in order to (1) decide the length of the survey interview; (2) improve the question wording, if necessary; (3) correct and improve translation; (4) delete unnecessary questions or add new ones, as required; (5) test the order of questions and identify bases; (6) check accuracy and adequacy of the questionnaire instructions; (7) catch any items which were conceptually vague; and (8) determine whether the questions were clear.

Workshops for area coordinators and field interviewers were conducted in strategic locations based on the sample spots that were generated. Workshop activity mainly consisted of one day of office training to learn the basics of the project. In the training, my research teams were presented with several guidelines. In particular, they were asked not to provide options 'don't know' or 'refuse to answer' to the respondents. These options were only for the enumerator to fill in privately if the respondents truly could or wouldn't respond with an answer. Importantly, during the workshop, enumerators were clearly instructed to conduct interviews in respondents' homes or other private and suitable places rather than to recruit them in public venues such as village offices or ward buildings. It was stated that there should be no village officials or heads of RT sitting in or listening when the interview took place. Given the unit of analysis of this study is an individual, other persons including household members were also not permitted to intervene when the interview was being conducted.

Having conducted the workshops, the data were then gathered through face-to-face interviews with randomly selected respondents. The area coordinators were responsible for obtaining necessary permission from local governments. Anchors reported to the field manager and observed interviewers, followed up, and did surprise checks on the field interviewers. We also ensured that logistics for field research were received quickly and administered correctly. One primary sampling unit (ten respondents) in Lhokseumawe, Aceh, was judged invalid for analysis because it contained fake interviews. In addition, it is important to note the number of calls and substitutions. If the field surveyors found selected respondents who couldn't be contacted during the first attempt, they were asked to visit a second time. My research teams successfully interviewed 1032 of the original respondents (85.3% of the total). A total of 178 respondents (14.7%) were unavailable to be contacted for various reasons, namely 'not accessible within the agreed upon field duration' (59.5%) (i.e. working out of the area or going to school outside the region); 'refused to be interviewed' (31.5%); 'being too old' (3.4%); 'very sick' (3.9%); and 'others' (1.7%). In these cases, we decided to substitute original respondents with others taken from another household beyond the covered intervals in the sample precinct.

More importantly, to make sure that the survey interview was actually carried out by the trained field researcher, we conducted layered spotchecks as quality control. The initial spot-checks were conducted randomly on 20% of the total sample by the persons in charge at the provincial level who returned to the selected respondents. The next level of spotchecks was done by national supervisors from the Jakarta office in the villages already checked by the area or provincial coordinators. In the quality control process, however, no significant error was found.

With regard to field editing, after each interview, the field researchers were asked to check their own work and check for consistency. All accomplished interview schedules were submitted to the assigned field anchor who, in turn, edited every interview. Once an incomplete or inconsistent answer was spotted in the questionnaire, the field interviewers were asked to revisit the respondent's house re-asking the question for verification. For purposes of the data processing, office editors checked for final consistency on all interviews before coding. A data entry computer program also verified the consistency of the encoded data before data tables and figures were generated. Having cleaned the questionnaires, 1210 (99%) of the postelection survey were judged valid for analysis. As noted above, although 178 out of 1210 respondents were substitutes, they possessed the same qualities (in terms of gender, age bracket, and socio-economic classes) as the original respondents. I was personally involved in all stages of the survey, from designing the survey, developing the survey instrument, training of field coordinators, and interviewers, pre-tests, spot-checks, cleaning, coding, and data entry.

To prove that my national public opinion survey was representative, we need to compare the demographic composition of the sample with the Indonesian population based on the most recent census (Table A.1). In terms of gender composition and urban-rural domicile, the sample of the survey was almost identical to the Indonesian population (see the left panel of Table A.1). This is largely because the population was initially stratified based on gender and based on area of residence. Interestingly, although the categories of religion and ethnic groups weren't determined from the outset, as shown in the right panel of Table A.1, the demographic samples based on these two important categories were similar to


**Table A.1** Demographic profiles of the sample compared to population (%)

the Indonesian population, confirming that the survey is representative of the country as a whole.

In addition, Table A.2 shows that survey sample demographics by province also reflect the Indonesian population. Since this survey unit is nationwide, the samples were drawn in proportion to the size of population of


**Table A.2** Sample demographics by province in comparison with the 2010 census (%)

each province. This is largely because each province has a different density of population. The higher population a province has, the more samples were generated. For instance, 17.5% of the sample was taken from the most densely populated province, West Java, compared to West Papua, accounting for only 0.4% of the total sample.

# Pre-election Surveys in January, February–March, and Late March 2014

In order to capture the level of intensification of vote buying efforts during the run-up to the 2014 election, I inserted a number of questions relating to this practice and other related issues into several Indikator national opinion surveys conducted in January, February–March, and late March 2014. All of these surveys applied standard survey methodology and procedures similar to those used for the post-election survey described above. They also implemented similar layers of quality control, from intensive workshops of interviewers to spot-checking, supervision, and data entry. Accordingly, this appendix simply describes the most important aspects of the survey methods, assuming that all steps involved in these projects were similar to what was already thoroughly explained in the preceding section.

First, face-to-face interviews were conducted during 18–30 January 2014. Multistage random sampling was used to produce a probability sample of 2039 voting-age adults drawn from all provinces in Indonesia and including an oversample from the Jakarta province. Based on this 2039-person sample, the overall margin of error reaches ±2.4% at 95% confidence level. There were 1678 original respondents (82.3%), while the substitute respondents were 261 people. Among those who couldn't be interviewed, 51% 'couldn't be contacted for various reasons' (i.e. working out of the area or at school outside the region), 14% 'weren't willing to be interviewed,' 4% was 'too old,' 2% was 'sick,' and so on. Quality control of the interview results was conducted randomly on 20% of the total sample by the supervisors who returned to the selected respondents. There were no significant errors found in this survey.

The second pre-election survey was conducted from 26 February to 6 March 2014. The sample size of 2050 is representative of the country as a whole. Based on the sample size, the estimated margin of error is ±2.2% at 95% confidence level. Selected respondents were interviewed through face-to-face interviews by trained interviewers. Of the 2050 cases in this survey, 1705 (83.2%) were original respondents, 345 cases involved respondent substitution. The third pre-election survey was conducted during 19–24 March 2014. The number of samples is 1220 voting-age adults who were selected with multistage random sampling proportionally distributed over the 34 provinces. The margin of error is around ±2.9% at 95% significance level. Of the 1220 cases in this survey, there were 1046 (85.6%) original respondents and 174 substitute respondents. Again, the modal answer to the question why original respondents couldn't be interviewed was 'inaccessible within the field duration' (57% and 51% in the February–March and the late March surveys, respectively). Notwithstanding there were around 15–16% of substitution, these substitute respondents generally had similar profiles to the original cases. In these surveys, quality control of the results was also carried out randomly on 20% of the total samples with strict supervision by the Indikator headquarter. No significant errors were found in either pre-election survey.

# Appendix B: Survey of Brokers and Candidates

In addition to relying on general population surveys, I also wanted to analyse the attitudes and experiences of both elected representatives and vote brokers with regard to vote buying. However, in conducting a survey which captured both groups, there were significant challenges, especially in designing a sample frame to select brokers. This is one of the biggest methodological problems I encountered: there is no ready-made sampling frame of vote brokers in Indonesia. Recent scholarship on vote brokers in Indonesia has greatly contributed to providing valuable insights into the underexplored nature of political brokerage (e.g. Aspinall, 2014a; Triantini, 2016), notably regarding its structure, networks, and varying modus operandi in distributing electoral handouts. However, little is known about the extent to which results from these convenience samples can be reliably projected to the whole population of vote brokers (Stokes et al., 2013: 261–262).

In order to address the problem of the absence of a sampling frame of brokers, and in an attempt to produce a sample that allows for a scientifically sound portrayal of the targeted population, this study generates a probability sample of brokers in a way inspired by Susan Stokes and her colleagues (2013), who collected information from randomly selected city councillors in Argentina. However, unlike Stokes and her colleagues, who considered the councillors as 'elected brokers,' my study doesn't refer to local parliament members—an entry point to provide the sampling frame—as political brokers. In addition, my study slightly contrasts with Stokes and her collaborators in terms of selection of non-elected brokers, as will be described below.

Given that I funded the construction of this unique dataset of brokers out of my own pocket, I was only able to draw a representative sample in four provinces. In order to determine the locations for the probability sample of brokers, unlike Stokes et al. (2013) who purposively selected four Argentine provinces from which to pick brokers, I was instead guided primarily by the fit of cases with the statistical evidence from a large dataset of electoral district surveys conducted by Indikator and SMRC during the run-up to the 2014 election. Using results from individual-level data, I measure the relationship between partisanship and vote buying as the determinant for case selection for both my MP and broker surveys (Fig. 1.2 in Chap. 1). In short, our massive dataset of pre-election surveys in 73 out of 77 electoral districts across Indonesia results in the following four provinces from which to sample MPs and brokers: Central Java, East Java, North Sulawesi, and West Sumatra. These areas not only reflect variations across the level of partisan closeness and vote buying but also vary with respect to socio-economic and political profiles, the topography of power, and the level of competitiveness (see Chap. 6 for more details).

The population of this survey is all DPRD (local parliament) members at both the provincial and district levels and their brokers, locally known as success team members, in four provinces (West Sumatra, Central Java, East Java, and North Sulawesi). Hence, it is worth noting that our data on candidates' behaviours and strategies in this study refer to newly elected local candidates (though of course some of them may have served previously in the DPRD). This is because when this unique, face-to-face survey was conducted from 30 September to 25 October 2014, these low-level politicians had been just elected in the April legislative elections and had been officially inaugurated as local MPs in August 2014. The sample was determined with multistage random sampling method. The size of the sample in each province was set at 300 respondents, so the sample in four provinces in total was expected to reach 1200 respondents (100 Provincial DPRD members, 300 campaign team members of Provincial DPRD members, 200 Regent/City DPRD members, and 600 success team members of Regent/City DPRD members). It is no exaggeration to say that our data are the largest research sample of political operatives in any region.1

<sup>1</sup>For comparative perspective, Stokes and her colleagues' survey in 2009 'only' generated a probability sample of 800 elected city councillors and non-elected brokers in four provinces in Argentina (Stokes et al., 2013).


**Table B.1** Population of local MPs in four provinces

First of all, I demonstrate the population of Provincial and District DPRD members in the four provinces from which the sample was drawn and then discuss the procedure of sample selection in a systematic way. As shown in Table B.1, in the four provinces, in total, there are 310 members in provincial legislatures (DPRD I) and 4210 members in district legislatures (DPRD II). Given that the number of district DPRD members is much greater than that of provincial DPRD members, and given this study is intended to capture intense lower-level campaigns, I decided to draw a larger sample of district DPRD members and their brokers than their counterparts from provincial legislatures.

Let's start with the sampling procedure for drawing representative samples of Provincial DPRD members and their brokers (Table B.2). First, the population in each province was grouped (stratified) according to its Provincial DPRD electoral districts. Second, in each Provincial DPRD electoral district, a sample of Provincial DPRD members was randomly selected based on proportion. Then, for every selected Provincial DPRD member, three of their brokers, who helped them during the last legislative election, were also randomly selected.2 As noted above, the population of number of brokers in each region was unknown. Accordingly, to determine the broker population, as already discussed in Chap. 6, and to draw a probability sample of brokers, I needed to collect the relevant data based on information mentioned by provincial MP respondents when the interviews were conducted.

<sup>2</sup>As I describe in the following pages, I divided the brokers into two categories: core team and base-level brokers. Hence, for every surveyed local Dewan Perwakilan Rakyat Daerah (DPRD) member, we then sampled one core team member and two base-level brokers.


**Table B.2** Population and sample sizes of provincial DPRD members and their brokers

In order to give a broader sense of this random selection, I will elaborate this detailed, complex procedure as follows. When the interview with each Provincial DPRD member was approaching the end, we asked:

How many brokers [*tim sukses*, *tim relawan*, *tim kampanye pemenangan*3] from the top level or core team [*tim inti*] to lowest-level member who deal directly with voters [door-to-door]—worked for you during the 2014 legislative election?

My field researchers then recorded this number. Note that district DPRD members were also presented with this question. Among those who responded to this question, the mean number of brokers working for provincial and district DPRD members was 149.44 with a standard deviation of 299.1. As already shown in Chap. 6, the number of brokers working for the sampled successful candidates varies across the province and varies in size, including one extreme outlier—one candidate who

<sup>3</sup> I used these three words for brokers in the survey instrument. As discussed in Chap. 6, a frequently mentioned expression for informal networks of vote brokers is 'success teams' (*tim sukses*). However, in some cases, people used a variety of names to call brokers, such as a less pejorative term, *tim relawan* whose closest English equivalent is 'volunteer team'; another neutral term is *tim kampanye pemenangan* (winning campaign team). In the Indonesian context, the word 'broker' has, in fact, been Indonesianised, but this word is considered by many as a more pejorative term than the often-used, neutral term *tim sukses*.

reported working with 3000 core and ground-level brokers. As previously detected by Stokes and her colleagues (2013: 267), low-level politicians are likely to inflate the number of brokers, but they may not be able to provide the names of their brokers. To address this problem, I first distinguish non-elected brokers into two categories: 'core team' (*tim inti*) and 'lowest-level brokers' or 'ground-level brokers,' whose main tasks were swaying voters in favour of the candidate they support. As Aspinall and his colleagues (2017: 10) explain, in Indonesia, this type of broker was known by a variety of terms, such as *kader* (or cadre) or *sabet* or *gapit* (Javanese terms meaning 'whip'). What concerned Stokes and her colleagues about the overestimating tendency above corresponds with my questionnaire pre-testing findings. The results from the pre-test on about five politicians and ten senior brokers show that my politician respondents often found it difficult to name their base-level brokers. They were able to answer how many brokers worked for them, but when asked to provide a list of names of their grassroots brokers, they typically said that it was their core campaign team (*tim inti*) who had the list. Then, we asked *tim inti* "how many base-level brokers did you have who were charged with influencing voters." The objective is clear: (1) to evaluate or match the numbers of the broker army given by each local MP member in the previous question. We found that most surveyed low-level politicians accurately reported the number of brokers who worked for them in 2014. (2) Given most *tim inti* had a complete list that contained the names and addresses of grassroots brokers, it was then instrumental to assess the probable number of brokers who worked in direct contact with voters.

In order to draw a representative sample of the core team, the sampled Provincial DPRD members were read the following prompt at the end of the interview:

We greatly appreciate your participation. The success of this study depends on your cooperation. We also need your help to choose some of your campaign team members for interviews. To ensure that we would interview a representative group of people, these people should be selected at random. Can you mention the names of the core campaign team who have helped you in the 2014 legislative election?

Interestingly, all the surveyed Provincial DPRD members were happily willing to accept the request and provided a bunch of names of their core teams, including their contact information. This is also the case for the sampled district MP members, as will be explained below. My researchers were then instructed to enter the list of names of *tim inti* in alphabetical order to select one respondent from the category of the core campaign team with simple random sampling. As noted above, the same procedure was done to determine the respondent category of the base-level brokers. When the interview with the sampled core team members was approaching the end, our research teams read the following prompt:

We greatly appreciate your participation. The success of this study depends on your cooperation. We also need your help choose the names of the campaign volunteer team [*tim relawan* or *tim sukses*] who worked at the very bottom level or in direct contact with the voters (door-to-door) for interviews. To ensure that we interviewed a representative group, these people should be selected at random.

Interestingly, most of the core team members handed over a printed list of their base-level brokers. Similarly, my field researchers then entered a list of names of the grassroots brokers provided by the surveyed core team member to the random sheet alphabetically to determine two respondents from the category of the base-level brokers who worked in direct contact with the voters. The sample in this stratum was selected by simple random sampling.

Having discussed the random selection of the sample of Provincial DPRD members and their brokers, I am now in a position to describe the sampling procedure for determining district local MP members and their political operatives. It was carried out in a generally similar fashion, but with a slightly different procedure: first, the population in each province was grouped (stratified) based on four different zones. Each zone was a combination of provincial electoral districts. I used specific criteria to establish such zones which vary depending on the geographical proximity, history, and conditions of a particular region. Second, in each zone, a regency/city (*kabupaten/kota*) was picked randomly as a sample by proportion. In total, ten regencies/cities were selected. Third, in each selected regency/city, five regency/city DPRD members were randomly selected as respondents (see Table B.3). Finally, for every surveyed district DPRD member, we sampled one core team member (*tim inti*) and two grassroots


**Table B.3** Population and sample sizes of district DPRD members and their brokers

brokers, using simple random sampling with the same procedure as explained above.

The survey instrument was written and administered in Indonesian. Because my study was exclusively conducted in Indonesia, it is difficult to make global comparisons. Accordingly, for international comparability, I used a number of questions that have been used elsewhere, including a 2009 broker survey by Stokes and her collaborators (2013). I also incorporated their innovative survey experiment in my brokers' survey (see Chap. 4).4 Needless to say, I also had my own particular questions related to the aspects of the Indonesian situation with which I was primarily concerned. Overall, as outlined in Chap. 1, with respect to the local politicians' survey, my survey instrument sought to collect information on targeting strategies, broker recruitment, the structure of their personal campaign team, voter pragmatism, the level of competitiveness, and broker monitoring. Regarding the brokers' survey, I asked a battery of questions about the mechanisms of vote buying (who is targeted and when, how, and why they are targeted), prior personal contact with candidates, rent-seeking behaviours, and so forth (see Chap. 6). As indicated above,

<sup>4</sup>One of the authors, Thad Dunning, posted the survey instrument in its entirety at http://thaddunning.com/data/brokers. Since the questionnaire was written in Spanish, I hired a trusted translator to translate it into Indonesian.

my survey instrument was piloted in late August 2014, three weeks prior to field implementation.

With the proposed 1200 respondents from provincial and district DPRD members and their brokers, this survey required at least 100 experienced, skilful interviewers. One field interviewer was assigned to interview one selected Provincial DPRD member, two selected District DPRD members, three provincial brokers, and six district brokers. In addition, I was helped by four area coordinators and their assistants, along with five national supervisors. I personally monitored the survey full-time. However, since the nature of this survey was completely different compared to voter surveys, I was first looking for not only competent but also experienced field researchers who had the ability to approach politicians in a way that would ensure they were willing to be interviewed. I personally recruited them based on input and discussion with my provincial coordinators. Most of them were senior researchers, had been working with my survey firm for a long time, and having Master's degrees, mostly in political and social sciences.

More importantly, prior to going to the field, intensive training of field researchers was conducted in four capitals of the provinces. I went to Padang (West Sumatra), Surabaya (East Java), Manado (North Sulawesi), and Semarang (Central Java) to directly train them in how to conduct face-to-face interviews with middle-level elites. One to two days of intensive workshops were conducted in each of these capitals to teach the basics and the raison d'être of the project and to understand the sampling procedure for determining a representative sample of local DPRD members and their brokers. The workshop was also intended to provide participants with appropriate communication skills to approach respondents and make them understand each question well before answering it.

To increase the response rate, I set a field period that would be lengthy enough for the field researchers to interview all the respondents assigned to them. Note that it isn't always easy to convince party elites to accept an interview request on sensitive questions like vote buying and its targeting strategies. Despite this practice being less likely to be stigmatised than in the past (see Chap. 2), it is still illegal in Indonesia. In some cases, especially when selected respondents were difficult to reach, I personally made a call to them, asking them politely on behalf of my interviewers to accept the interview request. On some occasions, I also accompanied members of my research teams to interview the respondents. Simultaneously, we also carried out layered spot-checks as quality control. The initial spot-checks were conducted by area coordinators in each province on 20% of the respondents under their supervision. The next level of spot-checks was carried out by national supervisors with 50% of the respondents already checked by provincial supervisors. We also used the telephone for callback and spot-checking: once we got the respondents' contact information from the survey form, my national supervisors rang them and cross-checked the information gathered during the interview. No significant errors were found.

Interestingly, notwithstanding the complexities of the sampling procedure and the nature of the survey, we got relatively high response rates. One sampled District DPRD member in one regency in East Java couldn't be interviewed within the agreed-upon field duration. Among 1199 respondents who participated in the survey, 1071 (89.3%) of the respondents were from our original selection. Only 128 (10.7%) were substituted. Reasons for respondent replacement varied: 46 out of 128 respondents (35.9%) were unwilling to be interviewed; 44 respondents (34.4%) were out of town; 34 respondents (26.6%) couldn't be contacted at all, or had an unclear address; and 1 respondent (0.8%) moved. We also decided to replace three respondents from West Sumatra because they claimed they didn't have base-level brokers. Because these substitute respondents generally had similar profiles to the original selections and the respondent substitution followed the procedure, I decided to include them in the analysis. Having cleaned the questionnaire, 1199 cases were judged valid for analysis. As indicated earlier, I was personally and directly involved in all steps of the project, from recruiting and training of the provincial coordinators and field researchers, pre-tests, spot-checks, cleaning, coding, and data input.

# Appendix C: Correlation Between Variables (Pearson's *r*)

© The Author(s) 2019 287 B. Muhtadi, *Vote Buying in Indonesia*, https://doi.org/10.1007/978-981-13-6779-3



\*\*\**p* < 0.001; \*\**p* < 0.01; \**p* < 0.05

# Appendix D: Normality Test of the Linear Regression Analysis of the Determinants of Vote Buying

Based on the results of the normality test below, the empirical distribution of the data (the histogram) in Model 1 isn't bell-shaped. Similarly, the histograms in both Model 2 and Model 3 don't resemble the normal distribution, suggesting that the residuals weren't normally distributed.

# Appendix E: Logistic Regression Analysis of the Determinants of Vote Buying

The dependent variable is a dichotomous measure that takes a value of 1 if the respondent received vote buying once/twice, several times, rarely, quite, or very often.



*b* = Unstandardised coefficients

Note: \**p* <0.05; \*\**p* < 0.01; \*\*\**p* < 0.001

# Appendix F: Multivariate Analyses of the Determinants of Winning and Losing Incumbents

This regression model designates all parties that participated in the previous elections—thereby having incumbent candidates running in 2014 and achieved parliamentary threshold in 2014 as the reference category: PKB, PKS, PDI-P, Golkar, Gerindra, Democrat, PAN, PPP, and Hanura.1 Each category in this variable has a value of 1 for its category and a 0 for all others. For example, in the first model (PKB), 1 = those who run from this party; 0 = otherwise. This is also true for other models.

1Therefore, I excluded candidates from NasDem because the party just ran for the first time in the 2014 elections.



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

# Bibliography

Books, Journal Articles, and Unpublished Papers


© The Author(s) 2019 297 B. Muhtadi, *Vote Buying in Indonesia*, https://doi.org/10.1007/978-981-13-6779-3

Andrews, J.T. and Inman, K. (2009) Explaining Vote Choice in Africa's Emerging Democracies. Paper presented at Midwest Political Science Association.

Aspinall, E. (2014a) When Brokers Betray: Social Networks and Electoral Politics in Indonesia. Critical Asian Studies 46(4): 545–570.

———. (2014b) Indonesia's 2014 Elections: Parliament and Patronage. Journal of Democracy 25(4): 96–110.

———. (2016) Don't You Forget About Me? Candidate Performance, Club Goods and Vote Buying in Indonesia. Paper presented at Biennial Conference of the Asian Studies Association of Australia (ASAA), 5–7 July, ANU, Canberra.


———. (2010) Swing Voters, Core Voters, and Distributive Politics. In: I. Shapiro, S. Stokes, E. Wood, and A. Kirshner (eds) Political Representation. Cambridge: Cambridge University Press, 342–357.


———. (2014b) Indonesia's Legislative Elections: The Importance of Money and Personalities. Australia Institute of International Affairs. Viewed at http:// www.internationalaffairs.org.au/australian\_outlook/indonesias-legislative -elections-the-importance-of-money-and-personalities/. Accessed 21 February 2017.


———. (2007a) How Do Rules and Institutions Encourage Vote Buying? In: F.C. Schaffer (ed) Elections for Sale: The Causes and Consequences of Vote Buying. Colorado: Lynne Rienner Publisher, Inc.

———. (2007b) How Effective are Institutional Reforms? In: F.C. Schaffer (ed) Elections for Sale: The Causes and Consequences of Vote Buying. Colorado: Lynne Rienner Publisher, Inc.

Hicken, A., Leider, S., Ravanilla, N., and Yang, D. (2017) Temptation in Vote-Selling: Evidence from a Field Experiment in the Philippines. Unpublished manuscript. http://www-personal.umich.edu/~leider/Papers/ TemptationVoteSelling.pdf. Accessed 16 May 2017.


———. (1996) Contested Meanings of Elections in the Philippines. In: R.H. Taylor (ed) The Politics of Elections in Southeast Asia. Cambridge: Cambridge University Press.


———. (2007) The Demise of Clientelism in Affluent Capitalist Democracies. In: H. Kitschelt and S. Wilkinson (eds) Patrons, Clients, and Policies: Patterns of Democratic Accountability and Political Competition. Cambridge, UK: Cambridge University Press.

Kitschelt, H. and Wilkinson, S. (2007a) Patrons, Clients, and Policies: Patterns of Democratic Accountability and Political Competition. Cambridge, UK: Cambridge University Press.

———. (2007b) Citizen-Politician Linkages: An Introduction. In: H. Kitschelt and S. Wilkinson (eds) Patrons, Clients, and Policies: Patterns of Democratic Accountability and Political Competition. Cambridge, UK: Cambridge University Press.


———. (2013) Vote Buying and Accountability in Democratic Africa. PhD dissertation, UCLA, CA, USA.


of Democratic Accountability and Political Competition. Cambridge, UK: Cambridge University Press.


———. (2007a) Is Vote Buying Undemocratic? In: F.C. Schaffer (ed) Elections for Sale: The Causes and Consequences of Vote Buying. Colorado: Lynne Rienner Publisher, Inc.

———. (2007b) Political Clientelism. In: C. Boix and S. Stokes (eds) The Oxford Handbook of Comparative Politics. New York: Oxford University Press.

———. (2009) Pork, by Any Other Name … Building a Conceptual Scheme of Distributive Politics. Paper presented at the American Political Science Association Conference, Toronto, Canada.


———. (1989) Interest in Politics. In: K. Jenings and J.W. Van Deth (eds) Continuities in Political Action. New York: Walter de Gruyter.


———. (1995b) Beyond SES: A Resource Model of Political Participation. American Political Science Review 89: 271–294.


#### Articles and Reports


# Index1

#### **A**

*Abangan*, 20, 173 Absolute, 6, 13, 37, 38, 109, 110, 129, 130, 143, 224, 244, 262 Additional document collection, 34, 37 Africa, 3, 63, 139 Afrobarometer, 3, 63 Agency loss, 16, 39, 134, 145, 153–156, 166, 171–206, 245, 262 Aliran, 19 Alms, 98, 162 Americas Barometer, 3, 63, 64n10 Amorphous, 15, 149–153, 262 Anecdotal evidence, 4, 232 Anxiety, 125, 146, 161, 164 Argentina, 4n2, 90n1, 135, 135n2, 138, 141, 142, 180, 200, 201, 252, 277, 278n2 Arisan, 181 Aspinall, E., 4, 6, 8, 11, 13, 16, 19, 20, 24, 27, 30n16, 35n18, 45n1, 57, 81, 98, 123, 124, 143, 144,

149, 157, 157n11, 158, 161, 162, 164, 174, 182–184, 186, 188, 192, 193n6, 196, 197, 204, 210, 214, 220, 222, 228, 229, 232, 233n10, 234, 250, 255, 270, 277, 281 Attachment, 5, 27, 92, 111, 114, 117, 128, 146, 247 Auction, 232 Authoritarian regime, 2, 3

#### **B**

Banjarese, 122, 123 Base areas, 35, 118, 137, 141, 142, 149, 157, 158, 162, 250, 252 Base voters, 15, 35, 61, 101, 198, 214 Basis, 14, 15, 24–26, 28, 32, 143, 144, 156n10, 157, 157n11, 158, 162, 183, 184, 196, 198n8, 199, 200, 205, 256, 269 BAWASLU, 223, 259 Benin, 3n1, 14n4, 64, 85 Birth place, 15, 157

1Note: Page numbers followed by 'n' refer to notes.

© The Author(s) 2019 311 B. Muhtadi, *Vote Buying in Indonesia*, https://doi.org/10.1007/978-981-13-6779-3

*Bisyarah*, 125 Brazil, 135n2, 138, 141, 142 Broker-level, 29 Broker predation, 10, 16, 155, 171, 196, 200, 206, 213, 220, 229, 245, 247 Brokers, 5, 45n2, 82, 109, 133, 171,

**C**

Candidate base, 118

209, 244


Comparative, 5, 11, 16, 22, 37, 39, 46, 64n10, 82, 109, 134, 138, 185, 204, 259 Compensation, 4, 163, 164, 204 Competition, 1, 15, 22, 24, 27, 35–37, 101, 126, 137, 143, 161, 165, 197, 215–217, 219, 221–226, 228, 233, 239, 244, 245, 249, 255, 257, 260 Competitive elections, 2, 17, 39, 57, 146, 155, 214, 219, 228, 229, 231, 239, 243, 250, 253, 254 Conditional loyalty, 126 Confusion, 16, 145, 156, 157, 245 Connections, 14–16, 34, 82, 133, 143, 144, 156–159, 167, 183, 184, 190, 191, 205, 245, 251, 252, 257, 261 Constituency size, 56, 60, 75, 192, 192n5, 196 Contingency, 8 Co-partisans, 6, 13, 15, 16, 25, 71, 118, 134n1, 138, 143, 155, 159, 165, 186, 205, 214, 215, 219, 221, 223, 228, 244 Core-voter model, 6, 9, 13, 14, 129, 135, 136, 143, 247, 249–252 Credibility, 2, 10, 20, 64, 247, 250, 251

#### **D**

*Daerah pemilihan*, 18 *Daftar Pemilih Tetap* (DPT), 148, 149 Dataset, 12, 16, 33, 56, 74, 76, 77, 89–92, 130, 181, 187, 197, 198, 235, 278 Dawn attack, 53, 231, 234 Deceive, 134, 245

Defect, 9, 10, 122, 135n2, 144, 164, 197, 246

Democratic Party, 20, 54, 55, 111n1, 116, 125, 130, 149, 200, 222, 224 Democratisation, 1, 2, 82 Determinants, 3, 5, 7, 12, 28, 77, 81–103, 200, 235, 237, 278,

291, 293–295


#### **E**

East Java, 27, 32, 34, 36, 73, 76, 117, 120, 125, 141, 152n9, 160, 172–174, 176, 182, 190, 192, 222, 223, 226, 278, 284, 285 Effectiveness, 3, 5, 7, 17, 39, 189, 192, 198, 200, 211, 216, 220, 233, 260 Effectiveness ratio, 7 Efficacy, 84–86, 88, 91–93, 100, 102 Electoral arena, 3 Electoral outcomes, 4, 11, 16, 27, 160, 214, 216, 229, 231, 235, 239, 246, 251, 253, 257

Electoral rules, 2, 18–28, 254 Electoral systems, 2, 3, 15, 17, 18, 22, 24–26, 134, 136, 166, 216, 226, 228, 239, 243, 249, 256–260 Endogeneity, 5, 10, 112, 113, 130, 247 Ethnic, 15, 16, 33, 90, 122, 143, 153, 158, 166, 187, 251, 272 Executive, 12, 29, 31, 38, 46, 47, 49, 55–61, 71–76, 116, 117, 222 Extent, 3, 5, 22, 25, 28–30, 32, 55, 71, 93, 111, 141, 156n10, 192, 214, 226, 253, 260, 277

#### **F**

Floating masses, 163 Focus group discussions (FGD), 34, 36, 196 Fraud, 2, 223 Free and fair, 1, 2

#### **G**

Generosity, 124, 158 Gerindra, 20, 27, 54, 111, 115, 116, 149, 222, 232, 295 Gift giving, 35, 74, 84, 98, 123–126, 162, 238 Gifts, 10–12, 16, 53, 54, 63, 66–68, 71–73, 76, 98, 109, 113, 118, 123–126, 128, 129, 131, 160, 162, 163, 166, 198, 199, 202, 210, 213, 220, 230–232, 236, 237, 251, 253 Golkar, 20, 22, 53, 54, 55, 70, 71, 111, 111n1, 115–117, 125, 130, 138, 149, 161, 162, 173, 221, 222, 233, 295 Goods, 4, 6, 8, 9, 14n4, 16, 20, 47–50, 55, 63, 63n9, 70, 75, 83, 84, 86, 100, 101, 116, 119,

123n3, 124, 125, 128, 133, 154, 163n15, 185, 188, 198, 201n9, 202, 209, 210, 225, 230, 233, 233n10, 234, 247, 253, 255 Grassroots, 11, 32, 36, 136, 142, 164, 174, 184, 186, 205, 261, 281,

282

#### **H**

Handouts, 4, 5, 7, 16, 32, 37, 55, 66, 68–70, 75, 76, 98, 100, 110, 113, 117, 123, 154, 193, 196, 202, 209–213, 220, 229, 233, 236, 238, 239, 246, 247, 251, 253–255, 262, 277 Hanura, 20, 53, 117, 123, 149, 152, 223, 295 Hicken, A., 8, 17, 22, 24, 26, 28, 30n16, 64, 82, 83, 87, 97, 144n4, 174, 192, 233, 234, 238, 261 Honduras, 69

#### **I**

Incentives, 2, 5, 9, 16, 17, 22, 25, 27, 37, 45, 50, 55, 55n6, 56, 57, 61, 63, 64n10, 67, 98n8, 99, 102, 113, 114, 126, 134, 143, 153, 155, 164, 165, 189, 197, 201, 209, 211, 215, 216, 219, 228, 229, 231, 233, 237–238, 244, 251, 254, 255, 257, 258 Income, 5, 83, 87, 89, 97, 97n7, 141, 164, 180, 182 Indifferent, 11, 84, 90n1, 115, 120, 160, 165, 211 Indikator, 30, 32, 71, 98n8, 145, 203, 225, 225n9, 254, 267, 267n1, 269, 274, 275, 278

Individual-level, 5, 9, 69, 81, 82, 94, 130, 278 Indonesian Survey Institute (LSI), 21, 26, 26n14, 30, 45n2, 46, 50n4, 55n6, 71, 72n16, 111n1, 145, 254, 257n6, 267, 267n1, 269 Inefficient delivery, 7 Informal, 11, 15, 34, 35, 83, 87, 91, 98, 137, 151, 156, 167, 173, 174, 230, 249, 261, 280n4 Informational model, 247, 247n1, 249, 251–253 Institutional changes, 2 Intensity, 7, 45, 46, 64, 68, 76, 93, 196 Internal party competitors, 13, 222 Interparty, 25, 226, 228, 230, 257 Interviews, 25, 29, 31, 32, 34, 35, 37, 54, 70, 71, 75, 76, 121–126, 128, 130, 144–146, 151, 154, 156, 158–160, 162, 174, 176, 182, 187, 188, 190, 191, 193, 194, 196, 199, 200, 204, 214, 231–234, 256, 262, 268–271, 274, 275, 279–282, 284, 285 Intraparty, 13, 15, 27, 126, 143, 152, 165, 217, 219, 221–225, 228, 239, 244, 245, 249, 255, 257, 260

#### **K**


Kramon, E., 8, 11, 52, 54, 65n11, 65n12, 66, 135, 210, 211, 243, 247, 247n2, 249–253, 260

#### **L**

Latin America, 63, 64, 86, 134–136, 135n2, 244, 250, 261 Latin American Public Opinion Project (LAPOP), 3, 4n2, 63, 64n10 Leakage, 6, 37, 39, 144, 154, 155, 171, 205, 246 Lebanon, 69 Legislative, 2, 11, 12, 16, 18, 18n5, 20, 21n9, 26, 27, 30, 32, 36–38, 45n2, 46–50, 53, 55, 55n6, 57, 59, 61, 63, 64, 75–77, 81, 91, 94, 97, 98n8, 101, 102, 110, 114–117, 119, 137, 147, 171, 175, 176, 184, 194, 197, 205, 211, 213, 214, 219–221, 225, 230, 236, 238, 257, 258, 267, 268, 278–281 Liberia, 64 List experiment, 29, 30, 53, 64–69, 65n12, 76, 260 Logics, 5, 7, 10, 11, 61, 101, 120, 121, 137, 157, 160, 163, 165, 205, 232, 247, 262 Loyal, 6, 10, 16, 84, 101, 111n1, 122–126, 129, 133, 134, 134n1, 143–147, 149–166, 183, 201, 209, 244, 245, 247, 251, 262 Loyalty, 14–17, 21, 34, 35, 38, 111–114, 117, 120, 122, 126, 134, 141, 143, 145, 146, 146n7, 148–153, 156–166, 196, 205, 245, 247, 251, 252, 256, 261

#### **M**

Maintenance buying, 9


#### **N**

NasDem, 20, 115, 152, 223, 226, 295n6 Neighbourhood question, 29, 76 Networks, 10, 11, 14–16, 14n4, 24, 33–35, 39, 83, 84, 87, 91, 91n3, 98, 99, 118, 133, 134, 137, 143–145, 153–160, 162, 165–167, 171–206, 234, 244, 245, 249–252, 260, 261, 277, 280n4 Nicaragua, 65n12, 69, 83 Nichter, S.C., 6, 8 North Sulawesi, 32, 117, 120, 152n9, 158, 172, 173, 176, 182, 192,

255, 278, 284

#### **O**

Obtrusive, 52, 69, 76, 197

Open-list proportional voting system, 3, 12, 243

#### **P**

*Pangsa pasar*, 122 Partai Amanat National (PAN, National Mandate Party), 19, 25n13, 27, 122, 125, 158, 194, 199, 224, 234, 295 Partai Demokrasi Indonesia Perjuangan (PDI-P), 19, 22, 25n13, 27, 53, 54, 55, 111, 111n1, 115–117, 130, 141, 149, 151, 152, 156n10, 173, 214, 215, 221, 252, 255 Partai Keadilan Sejahtera (PKS, Prosperous Justice Party), 19, 25n13, 115, 117, 146, 152, 223, 295 Partai Kebangkitan Bangsa (PKB, National Awakening Party), 19, 27, 122–124, 126, 141, 146,

154, 155, 156n10, 157, 158, 204, 223, 252, 295 Partai Persatuan Pembangunan (PPP, United Development Party), 19, 117, 122, 145, 182, 191, 223, 227, 295 Partisan preferences, 14, 252 Partisan proximity, 11, 247 Partisanship, 5, 12, 13, 32, 32n17, 34, 38, 99, 101, 103, 110–117, 126–128, 130, 133, 134, 138, 138n3, 139, 141, 145, 146n7, 147, 172, 244, 249, 256, 261, 278 Partisan voters, 13, 14, 37, 38, 99, 102, 103, 126, 128–130, 133, 134, 137, 141–143, 145–147, 152, 152n9, 153, 164, 165, 244, 245, 252, 262 Party base, 12, 34, 38, 39, 103, 116–120, 133, 134, 142–144, 151, 152, 156, 159, 166, 205, 249 Party-centred campaigns, 2 Party ID, 4n3, 5, 12, 17, 19, 22, 27–28, 30, 32, 38, 81, 84, 88, 89, 90n1, 91, 92, 94, 99, 101–103, 109, 110, 112–117, 130, 138, 139, 141, 256 Party loyalists, 6, 10, 11, 13, 14, 34, 37–39, 101, 103, 109, 110, 120, 121, 123, 125–128, 130, 133, 141–143, 145, 147–149, 152, 152n9, 165, 171, 196, 211, 244, 245, 262 Party membership, 21, 21n9, 22, 28, 30 Party system, 2, 18–28, 141, 250, 255n3 Party vote, 2, 23–25, 225n8, 254 Patronage, 3, 10, 12, 15, 16, 20–22, 27, 45, 46, 55, 61, 70, 84, 87, 103, 118, 120, 126, 143, 156,

158–160, 160n13, 162, 166, 185, 188, 198, 233n10, 234, 243, 245, 247, 251, 253, 254 Patterns, 3, 4, 7, 11, 15, 17, 24, 27, 29, 31, 33, 36, 38, 59, 64, 83, 85, 88, 103, 110, 111, 128, 130, 143, 157, 166, 172, 184, 192, 196, 198, 214, 219, 224, 229, 236, 247, 260 Personality, 24, 187, 257 Personal loyalist model, 247, 249, 250 Personal networks, 14–16, 14n4, 38, 39, 118, 133, 134, 137, 143–145, 154–160, 165, 166, 171–206, 252 Personal ties, 133, 134, 155, 159, 191, 196 Personal vote, 2, 6, 12, 13, 15, 24–27, 25n11, 25n13, 134n1, 137, 142–144, 154, 159, 205, 214–216, 218, 219, 224, 225n8, 226, 228, 231, 233, 239, 244, 245, 250, 254, 257 Persuasion, 9, 198 Perverse accountability, 86, 204, 259 *Pesantren*, 98 Philippines, 4, 14n4, 30n16, 53, 63, 71, 83, 85, 124, 261 Pocket money, 74, 124 Policy positions, 2, 20, 20n8, 137, 256 Politeness, 124 Political attitudes, 5, 82, 84–89, 91, 97, 99–102 Political information, 84, 85, 91, 93, 103 Political interest, 84, 85, 88, 91, 93, 99, 100, 102 *Politik uang*, 4 Predictor, 5, 13, 15, 32, 38, 100, 101, 109, 112 Principal-agent problems, 6, 171, 262

Prisoner's dilemma, 232, 246 Puzzles, 5, 6, 103, 110, 128, 131, 133, 142–156, 165, 206, 209, 213, 220, 229, 238

#### **Q**

Qualitative approaches, 4, 28, 33–37, 260 Quiet period, 194

#### **R**

Reciprocate, 82, 84, 118, 124, 145, 154–156, 165, 166, 190, 199, 200, 205, 213, 244, 251, 253 Reciprocity, 84, 124, 135, 247, 247n1, 249–251, 253 Relative, 6, 17, 25, 33, 37, 38, 52, 55, 56, 69, 73, 90, 91, 109, 110, 127, 128, 130, 138, 172, 174, 189, 192, 203, 213, 215, 236, 238, 239, 246, 262 Relative risk, 126 Religious gatherings, 98 Religious networks, 15, 158 Rent-seeking, 10, 16, 39, 171, 186, 196, 198, 198n8, 200, 205, 209, 245, 283 Responsiveness, 124 Rewarding loyalists, 9 Risk aversion, 9, 13, 120–123, 130, 247

#### **S**

Saiful Mujani Research and Consulting (SMRC), 30, 32, 45n2, 50n4, 71, 225n9, 267n1, 278 Sampling frame, 31, 149, 277 *Santri*, 122, 173

Scope, 3, 4, 7, 28, 260

Secret ballot, 6 Secure, 23, 24, 29, 37, 122, 123, 125, 159, 162, 193, 196, 214, 231, 232, 239, 258, 260 Single-member district electoral system, 258 Slippage, 144, 191, 205 Social desirability, 28, 30, 46, 49n3, 64–71, 192, 194, 197, 198n8, 200–201, 211 Social norms, 84, 124, 125 Social stigma, 28, 64 Souvenir, 75, 124 Stokes, S.C, 5, 6, 9–11, 14, 192, 196, 200, 201, 204, 247, 249, 250, 253, 259, 277, 278, 278n2, 281, 283 Success team, 47, 48, 63, 119, 121, 125, 137, 150, 152, 172, 182, 184–188, 197, 198n8, 199, 200, 202, 203, 205, 212, 225, 230, 232, 255, 278, 280n4 Survey, 45, 89, 110, 145, 171, 172, 211, 280n4

Swing-voter approach, 6

#### **T**

Targeting mechanisms, 3, 5, 7, 11 *Tembakan terakhir*, 125 Trust, 29, 84, 86, 88, 91, 93–94, 102, 159 Turnout, 4, 7, 9, 10, 13, 16, 39, 84, 94, 99, 121, 137, 146, 209–212, 229, 238, 239, 247, 249–253, 260 Turnout buying, 9, 39, 119, 135, 135n2, 137, 138, 141–143, 154, 211, 244, 250

#### **U**

Uganda, 64 Uncertain, 123, 198, 210, 216, 233, 262 Uncommitted voters, 6, 13, 17, 118, 121, 122, 127–130, 145, 155, 156, 162, 163, 171, 198, 199, 211, 245, 247, 251, 262 Unreliable, 123, 171, 262

# **V**

Vote buying, 1–9, 11–14, 14n4, 16–18, 27–39, 32n17, 45–77, 81–103, 109, 110, 112–117, 121, 123–131, 133–138, 142–145, 151, 154–156, 159, 161, 163, 165–167, 171, 172, 174, 188, 189, 191–194, 196, 198, 204–206, 209–239, 243–263, 270, 274, 277, 278, 283, 284, 291, 293–294 Vote choice, 4, 12, 16, 37, 39, 201, 203, 211–213, 219, 229, 238, 239, 246, 249, 250, 252, 253 Voter compliance, 155, 156, 229 Voter intimidation, 234

#### **W**

