DIVERSITY, PATRONAGE AND PARTIES: PARTIES AND PARTY SYSTEM CHANGE IN INDONESIA by NATHAN WALLACE ALLEN A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Political Science) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) December 2012 ©Nathan Wallace Allen, 2012 ii Abstract The dissertation asks why some political systems have many parties and others only a few. Existing research on regime survival confirms that an excess of parties can generate regime instability. Because party systems affect key political outcomes, institutional engineers have sought to tweak electoral systems in order to produce favourable patterns of political competition. In particular, institutional designers in diverse states have attempted to curtail party system expansion by banning ethnically-based parties. Despite the growing popularity of such techniques, analysts know little about how these institutions work in practice. To address this issue, the dissertation explores a puzzle from the Indonesian case: the number of effective political parties in an electoral district strongly correlates with ethnic diversity, yet there is a de-facto prohibition of ethnic parties. Established theories linking ethnic diversity and party system size assume both the existence of ethnic parties and clear patterns of ethnic voting. However, neither one is present in the Indonesian case. The dissertation demonstrates that ethnic diversity has an indirect effect on party systems. It generates sub-national rent-seeking opportunities, a combination of high state involvement in the economy and weak constraints prohibiting the abuse of state resources for personal and political gain. In diverse electoral districts, the livelihoods of voters and elites are tightly linked to the control of the state. Elites have more opportunities to engage in rent-seeking behavior, affecting the way they participate in the political sphere. First, the opportunity to manipulate state resources draws elites into the electoral arena, increasing the number of viable candidates. Second, the intense focus on local goods distribution diminishes the value of national party platforms, allowing candidates to pursue political office under minor party labels. Third, voter demands for particularistic goods distribution lead them to disregard party labels and form tight patron-client linkages with candidates. The electoral consequence of these three phenomena is the expansion of the vote attained by minor parties, which act as vehicles of convenience for locally oriented rent-seeking networks. In high diversity / high rent electoral districts, the expansion of the vote attained by the minor parties fragments the party system. iii Table of Contents Abstract.............................................................................................................................. ii Table of Contents ............................................................................................................. iii List of Tables ................................................................................................................... vii List of Figures................................................................................................................... ix Glossary ............................................................................................................................ xi Acknowledgements ........................................................................................................ xiii Chapter 1 - Introduction .................................................................................................. 1 The puzzle of party system size in Indonesia.............................................................................. 4 Theory......................................................................................................................................... 6 Mechanism................................................................................................................................ 10 The importance of the Indonesian case..................................................................................... 11 Contributions............................................................................................................................. 13 Plan of dissertation.................................................................................................................... 14 Chapter 2 – Theory......................................................................................................... 16 Electoral fragmentation in Indonesia ........................................................................................ 16 Explaining electoral fragmentation: state of the literature ........................................................ 17 Electoral institutions ............................................................................................................................18 Social cleavages ...................................................................................................................................20 The interactive synthesis ......................................................................................................................22 Party aggregation .................................................................................................................................25 A theory of party system size across electoral districts ............................................................ 28 Defining rent opportunities ..................................................................................................................29 Rents and candidates strategies: entry, affiliation, and campaigns ......................................................31 Rent opportunities and party system size .............................................................................................35 Scope conditions and contributions .......................................................................................... 36 Rent sharing expectations ....................................................................................................................36 Ethnic parties and ethnic party bans.....................................................................................................38 Decentralization and sub-national power .............................................................................................39 Chapter 3 - Indonesia Background ............................................................................... 41 The origins of nationalist norms ............................................................................................... 41 Ethnicity in Indonesia ..........................................................................................................................42 The rise of Indonesian nationalism ......................................................................................................44 Indonesian political institutions ................................................................................................ 46 Executive-legislative structure .............................................................................................................46 Electoral systems and party laws .........................................................................................................51 Indonesian party system: past and present ................................................................................ 57 iv The partisan actors ...............................................................................................................................57 Parties and cleavage structure ..............................................................................................................61 Gotong Royong and the power sharing tradition..................................................................................66 Rent opportunities and decentralization.................................................................................... 68 Decentralizing Indonesia......................................................................................................................69 Rent opportunities and ethnic diversity................................................................................................74 Diversity and rent opportunities: explaining the relationship ..............................................................79 Conclusion ................................................................................................................................ 85 Chapter 4 - Candidate Entry in Indonesia ................................................................... 87 Why enter? Existing explanations............................................................................................. 89 Empirical puzzle ..................................................................................................................................89 Approaches to entry .............................................................................................................................90 Ethnicity and entry ...............................................................................................................................92 Rent opportunities and entry ..................................................................................................... 94 Rents and diversity...............................................................................................................................94 Office benefits, party influence, and entry ...........................................................................................95 Case background: selecting candidates in Indonesia ................................................................ 97 Partisan context ....................................................................................................................................97 Campaigns and partisan influence .......................................................................................................98 Theory testing: entry across electoral districts........................................................................ 101 Ethnic diversity and entry ..................................................................................................................101 Rent opportunities and entry ..............................................................................................................106 Alternative explanations ......................................................................................................... 110 Communal voting...............................................................................................................................110 Strategic parties..................................................................................................................................111 Conclusion .............................................................................................................................. 114 Chapter 5 - Patronage and Party Switching in Indonesia ........................................ 116 Party affiliation: existing explanations ................................................................................... 118 Party switching in the literature .........................................................................................................118 Affiliation and candidate goals: local vs. national strategies .............................................................119 Party affiliation and rent opportunities ................................................................................... 120 Local rent opportunities and career strategies: a party switching hypothesis ....................................121 Case background: selection processes in Indonesia................................................................ 122 Selection processes: the party’s view.................................................................................................122 Selection processes: the candidate’s view..........................................................................................125 A party affiliation dataset........................................................................................................ 129 Generating the dataset ........................................................................................................................129 Describing the data.............................................................................................................................131 Validity testing...................................................................................................................................132 Theory testing: career strategies across districts ..................................................................... 136 Hypothesis 1: number of switchers across districts............................................................................136 Hypothesis 2: number of loyalists across districts .............................................................................138 Hypothesis 3: relative proportion of switchers across districts ..........................................................138 vAlternative explanation: switching and ethnicity ...............................................................................141 Conclusion .............................................................................................................................. 142 Chapter 6: Patronage, Ethnicity and the Personal Vote in Indonesia ..................... 144 Personal voting: existing explanations.................................................................................... 145 Defining the personal vote .................................................................................................................145 Explaining the personal vote ..............................................................................................................146 Ethnic diversity and personal voting ..................................................................................................148 Rent opportunities and personal voting .................................................................................. 149 Vote buying, time horizons, and the personal vote ............................................................................150 Rent opportunities and clientelist exchange .......................................................................................151 Case background: campaigns in Indonesia ............................................................................. 152 Monitoring and punishment ...............................................................................................................153 Decentralization and clientelism ........................................................................................................155 Theory testing: personal voting across electoral districts ....................................................... 156 Alternative explanation: ethnic diversity and personal voting................................................ 165 Ethnic diversity and candidate campaigns .........................................................................................166 Ethnicity and public opinion ..............................................................................................................171 Communal voting revisited ................................................................................................................176 Conclusion .............................................................................................................................. 177 Chapter 7 – Party System Size..................................................................................... 179 Ethnic diversity and party systems: existing explanations...................................................... 180 The curious treatment of ethnic parties ..............................................................................................180 Ethnic identities and voting................................................................................................................183 Party systems and diversity: mechanisms reconsidered .....................................................................184 The Indonesian context ......................................................................................................................185 Rent opportunities, ethnic diversity, and party system size .................................................... 186 Hypotheses .........................................................................................................................................187 Theory testing: district level party systems............................................................................. 189 Hypotheses 1 and 2: effective number of electoral parties.................................................................189 Hypothesis 3: the effect of rent opportunities in the absence of ethnic diversity ...............................191 Hypothesis 4: the effect of ethnic diversity in the absence of rent opportunities ...............................197 Conclusion .............................................................................................................................. 200 Chapter 8 – Party System Change .............................................................................. 201 Indonesian party system change: existing explanations.......................................................... 202 Electoral institutions and ethnic diversity: the dogs that did not bark................................................202 Presidentialism and new issues: explaining change ...........................................................................204 Rent opportunities, rent sharing, and party system change..................................................... 205 Machine politics or partisan melee? The effect of rent sharing expectations.....................................206 Rent sharing in post-Suharto Indonesia .............................................................................................207 Theory testing: change across time ......................................................................................... 208 Hypothesis 1: municipal party system size in 1999 ...........................................................................208 Hypothesis 2: district level party system change over time ...............................................................211 vi Hypothesis 3: The decline of Golkar and party system change .........................................................215 Conclusion .............................................................................................................................. 224 Chapter 9 - Conclusion................................................................................................. 225 Research findings.................................................................................................................... 225 Theoretical contributions ........................................................................................................ 229 Extensions and future research ............................................................................................... 232 Discussion of policy implications ........................................................................................... 234 Bibliography .................................................................................................................. 237 Appendices..................................................................................................................... 250 Appendix A – Map of Indonesia............................................................................................. 250 Appendix B – Supplement to Chapter 2 ................................................................................. 251 Appendix C – Supplement to Chapter 3 ................................................................................. 256 Appendix D – Supplement to Chapter 4 ................................................................................. 268 Appendix E – Supplement to Chapter 5.................................................................................. 274 Appendix F – Supplement to Chapter 6.................................................................................. 278 Appendix G – Supplement to Chapter 7 ................................................................................. 288 Appendix H – Supplement to Chapter 8 ................................................................................. 298 Appendix I – List of interviews .............................................................................................. 305 vii List of Tables Table 1 – Party system size in national electoral districts: descriptive statistics .............. 17 Table 2 - Party system size: theoretical predictions.......................................................... 38 Table 3 – Major party performance .................................................................................. 58 Table 4 – Candidate entry ............................................................................................... 105 Table 5 – Switching and loyalty: descriptive statistics ................................................... 132 Table 6 – Career strategies across time in major parties ................................................ 133 Table 7 – Switching by party dyad ................................................................................. 136 Table 8 – Switching across districts................................................................................ 137 Table 9 – Loyalty across districts ................................................................................... 138 Table 10 – Proportion of switchers across districts ........................................................ 139 Table 11 – Ethnicity and switching across districts ........................................................ 141 Table 12 – Rent opportunities and preference voting ..................................................... 161 Table 13 – Predicated impact of rents on preference voting .......................................... 162 Table 14 – Public preferences for co-ethnic leadership.................................................. 174 Table 15 – Determinants of party system size ................................................................ 190 Table 16 – Municipal level party system size................................................................. 194 Table 17 – Predicted values of municipal party systems ................................................ 195 Table 18 – Determinants of DPD electoral fragmentation ............................................. 198 Table 19 – Party system size [1999] ............................................................................... 209 Table 20 – Predicted municipal party system size [1999] .............................................. 211 Table 21 –Determinants of party system expansion ....................................................... 213 Table 22 – Predicted party system expansion................................................................. 215 Table 23 – Golkar support by election............................................................................ 222 Table 24 - Appendix B – Party system inflation example .............................................. 252 Table 25 - Appendix B – Determinants of party system inflation (national) ................. 253 Table 26 - Appendix B – Determinants of party system inflation (municipal) .............. 254 Table 27 - Appendix C – Determinants of civil service size .......................................... 257 Table 28 - Appendix C – Transfers, diversity, and population size ............................... 260 Table 29 - Appendix C – Determinants of corruption .................................................... 262 viii Table 30 - Appendix C – Determinants of infrastructure provision [2007] ................... 266 Table 31 - Appendix C - Determinants of infrastructure provision [2003] .................... 267 Table 32 - Appendix D – Determinants of entry [1999-2009] ....................................... 270 Table 33 - Appendix D – Determinants of provincial entry ........................................... 271 Table 34 - Appendix D – Determinants of upper house entry........................................ 271 Table 35 - Appendix D – Determinants of entry by party size ....................................... 272 Table 36 - Appendix D – Entry across levels of governance ......................................... 273 Table 37 - Appendix E – An interactive model of party switching ................................ 276 Table 38 – Appendix F – Transfers and preference voting ............................................ 278 Table 39 - Appendix F –Fixed effects model of preference voting ................................ 279 Table 40 - Appendix F – Effect of rents by party ........................................................... 280 Table 41 - Appendix F – Preference voting and district population size........................ 282 Table 42 - Appendix F - Voting and provincial population size .................................... 283 Table 43 - Appendix F – Preference voting and separatism........................................... 284 Table 44 – Appendix G – Ethnic diversity of party system size [2004-2009] ............... 288 Table 45 - Appendix G – Rents and party system size [2004] ....................................... 289 Table 46 - Appendix G – Alternative specification of rents [2004] ............................... 290 Table 47 - Appendix G - Alternative specification of rents............................................ 290 Table 48 - Appendix G – Determinants of municipal party system size ........................ 294 Table 49 - Appendix G – Alternative specifications of rents (municipal)...................... 295 Table 50 - Appendix G – Determinants of upper house fragmentation.......................... 296 Table 51 - Appendix H – Determinants of party system size [1999] ............................. 298 Table 52 - Appendix H – Determinants of municipal party system size [1999] ............ 299 Table 53 - Appendix H – Determinants of party system expansion ............................... 302 Table 54 - Appendix H – Party change and party expansion ......................................... 303 Table 55 - Appendix H – Determinants of Golkar support ............................................ 304 Table 56 - Appendix I – List of interviews..................................................................... 305 ix List of Figures Figure 1 - Electoral parties and district magnitude ........................................................... 20 Figure 2 - Electoral parties and ethnic fractionalization ................................................... 24 Figure 3 – Party system aggregation................................................................................. 27 Figure 4 - Theory mapping ............................................................................................... 31 Figure 5 – Ethnic diversity and civil service size ............................................................. 76 Figure 6 – Ethnic diversity and corruption ....................................................................... 77 Figure 7 – Ethnic diversity and infrastructure quality ...................................................... 79 Figure 8 – Candidates-per-seat in 2004 ............................................................................ 89 Figure 9 – Candidates-per-seat in 2009 ............................................................................ 90 Figure 10 – PAN candidates running in harness in Medan ............................................ 100 Figure 11 – PPNUI candidates running in harness in Medan ......................................... 100 Figure 12 – District level diversity and infrastructure quality ........................................ 107 Figure 13 – Candidates-per-seat by infrastructure quality.............................................. 109 Figure 14 – DPD entry across provinces ........................................................................ 111 Figure 15 - Loyalty across time in major parties ........................................................... 135 Figure 16 – Switching across districts [2004-2009] ....................................................... 140 Figure 17 – Preference voting and rents ......................................................................... 158 Figure 18 – DPR candidate in Karo ................................................................................ 167 Figure 19 – DPR candidate in Pemantang Siantar.......................................................... 168 Figure 20 – DPR candidate Sembiring in Karo .............................................................. 169 Figure 21 - Sembiring in Pematang Siantar .................................................................... 169 Figure 22 – DPD candidate Batubara in Karo ................................................................ 169 Figure 23 – Batubara in Medan ...................................................................................... 169 Figure 24 – DPRD candidate Sumer in Bali ................................................................... 170 Figure 25 – Public preferences for co-ethnic leadership across provinces ..................... 176 Figure 26 – Marginal effect of ethnic diversity on party system size............................. 196 Figure 27 - Ethnic diversity and electoral fragmentation in DPD elections ................... 198 Figure 28 – Rent opportunities and party system expansion .......................................... 214 Figure 29 – Golkar support by region............................................................................. 218 xFigure 30 – Golkar’s decline........................................................................................... 223 Figure 31 - Appendix A – Map of Indonesia .................................................................. 250 Figure 32 - Appendix B – Municipal party system inflation .......................................... 255 Figure 33 - Appendix C – Civil service size and transfers ............................................. 259 Figure 34 - Appendix E – Marginal effect of corruption on party switching ................. 277 Figure 35 - Appendix G – Minor party support and electoral fragmentation [2004] ..... 291 Figure 36 - Appendix G - Minor party support and electoral fragmentation [2009] ...... 292 Figure 37 - Appendix G - Minor party support and municipal fragmentation ............... 292 Figure 38 - Appendix G – Straight ticket voting ............................................................ 293 Figure 39 - Appendix G – Upper house fragmentation and candidate numbers ............ 297 Figure 40 - Appendix H – Marginal effect of diversity on electoral fragmentation....... 300 Figure 41 - Appendix H – Party system size 1999-2009 ................................................ 301 Figure 42 - Appendix H - Party system expansion ......................................................... 301 xi Glossary BPS Badan Pusat Statistik (Central Statistics Agency) CSIS Centre for Strategic and International Studies; Jakarta-based think- tank DAU Dana Alokasi Umum (General Allocation Fund); central government transfer to sub-national bodies DAK Dana Alokasi Khasus (Special Allocation Fund); central government transfer to sub-national bodies DPD Dewan Perwakilan Dareah (Regional Representative’s Council); elected national regional advisory council DPR Dewan Perwakilan Rakyat (People’s Representative Council); national legislature DPRD Dewan Perwakilan Rakyat Daerah (Regional People’s Representative Council); Provincial legislature DPRDII Dewan Perwakilan Rakyat Daerah (Regional People’s Representative Council); municipal legislature ENEP Effective number of electoral parties Gerindra Partai Gerakan Indonesia Raya (The Greater Indonesia Movement Party) Golkar Partai Golongon Karya (Functional Groups Party); governing party during the New Order Gotong Royong Term for mutual help or cooperation for shared benefit Gurem Term for small “chicken-flea” parties Hanura Partai Hati Nurani Rakyat (People's Conscience Party) Kabupaten/kota Sub-provincial administrative units; referred to as “municipalities” in the dissertation KPPOD Komite Pemantuan Pelaksanaan Otonmi Daerah (Regional Autonomy Watch); Indonesian NGO KPU Komisi Pemilu Umum (General Elections Commission) Masyumi Majelis Syuro Muslimin Indonesia (Indonesian Muslim Advisory Council); modernist Muslim party MPR Majelis Permusyawaratan Rakyat (People’s Consultative Assembly); supra-legislative body NU Nahdlatul Ulama (Revival of Islamic Scholars); traditionalist religious organization, ran as a party in 1955 elections Pemekaran Lit. “Blossoming”; term for division of sub-national administrative units PAN Partai Amanat Nasional (National Mandate Party) PBB Partai Bulan Bintang (Crescent Star Party) PBR Partai Bintang Reformasi (Reform Star Party) PD Partai Demokrat (Democratic Party) PDI-P Partai Demokrasi Indonesia Perjuangan (Indonesian Democratic Party of Struggle) xii PDK Partai Persatuan Demokrasi Kebangsaan (United Democratic Nationhood Party) PDP Partai Demokrasi Pembaruan (Democratic Renewal Party) PKB Partai Kebangkitan Bangsa (National Awakening Party) PKI Partai Komunis Indonesia (Indonesian Communist Party) PKS Partai Keadilan Sejahtera (Prosperous Justice Party) PNI Partai Nasionalis Indonesia (Indonesian Nationalist Party) PPP Partai Persatuan Pembangunan (United Development Party) PR Proportional Representation suku Sub-national identity group; used synonymously with ethnicity in Indonesia TI Transparency International UUD ’45 Undang-Undang Dasar 1945 (Basic Law of 1945); Indonesia’s first constitution xiii Acknowledgements I have accumulated many intellectual and personal debts while writing this dissertation (a few financial ones as well). Benjamin Nyblade went above and beyond in supervising this project. He has uncovered hidden assumptions, pointed out logical inconsistencies, recommended new literatures, set firm deadlines, encouraged ideas that worked, discarded ideas that failed, and found all important income streams to keep a roof over my head. Diane Mauzy has been keeping me honest in my interpretation of Indonesian politics since I first started writing on the subject. Ken Carty has had continuous advice on how to make the rather anomalous politics of Indonesia relate to the broader political parties literature. The feedback I have received from my committee has been invaluable and I will be forever grateful for their intellectual generosity. This project has benefitted from engagement with a long list of colleagues from the department. I have been lucky to have around political science students who shared my regional interests. Shane Barter has not only been a constant source of advice, he has also shared his contacts and electoral data. Aim Sinpeng has always been up for a conversation about Southeast Asian politics, and Netina Tan has brought the analytical hammer down more than once, always to the benefit of the project. I also need to thank Go Murakami, who I have run to on more than one occasion for statistical advice. Faculty members have provided invaluable support. On more than one occasion comments from Alan Jacobs and Richard Johnston prompted me to re-think my approach to an issue. Campbell Sharman has suggested new ways to think about exchange politics. Thanks as well to Chris Kam, who allowed me free use of his “super computer” when my humble laptop was not up to the computational tasks demanded of the dissertation. This project has also benefited from countless informal conversations with colleagues, especially Bill, Adam, Michael, Daniel, John, and Clare. Completion of this project was only made possible because of the support I received from friends and colleagues in Indonesia. Thanks to Yuli, for being a constant companion during both trips. I loved being able to get a slice of “normal” life in Jakarta. Rahdian had xiv the unenviable task of improving my Bahasa. He also provided insightful thoughts on Indonesian culture, and I greatly valued the time I was able to spend with both him and Desmond. They introduced me to Inge, who proved an invaluable research companion in Jakarta. My time in North Sumatra proved productive mainly due to the hard work of Ines. I learned much from my conversations with Husnul and his colleagues at USU. Key datasets were generously provided by Sunny and Benny. Thank you to all my friends and family who have patiently supported me as I completed this dissertation. Doll and Bruce provided a family-away-from-home for my first years in Vancouver. My parents and grandparents have served double duty as both cheerleaders and, in a pinch, lenders of last resort. They can all breathe a sigh of relief now that my educational odyssey is at last complete (sort of). Finally, thank you to Jill. I suspect she had no idea what she was getting into when she started dating a PhD student. But she has borne the hardships and in the midst of all of this educational craziness we have managed to build a loving relationship. And now that we have survived work-induced stress, poverty, extended absence, exhaustion, more stress, pressing deadlines, and more poverty, married life will seem easy, right? 1Chapter 1 - Introduction Why do some political systems have many parties while others have only a few? The origin of party system size is one of the oldest topics in modern comparative political science.1 Scholars continually return to the topic because they believe the number of parties in a system has consequences. While a system with a high number of parties is associated with positive outcomes, like a wider spectrum of ideological representation and increased voter affect, there are considerable downsides.2 Most ominously, a fragmented party system is associated with the prevalence of extremism and polarization. In the infamous examples of 20th century Europe, party system fragmentation contributed to debilitating deadlock in France and the fall of the democratic regime in Germany (Sartori 1966). Cross-national research on regime survival confirms that an excess of parties is dangerous to the stability of democratic regimes.3 Current literature indicates that party system size is shaped by the electoral system, the presence of societal cleavages, and the relative centralization of power in a political system. Two of these factors are relatively fixed: modifying the relative centralization of power most often requires lengthy constitutional change while deliberate alterations of demographic structures take place over the long-term (at least in liberal democratic states). Electoral systems, on the other hand, can typically be manipulated through simple acts of the legislature. Because party systems directly affect outcomes people care about, institutional engineers have sought to tweak electoral systems in order to produce favourable patterns of political competition (Horowitz 2003; Lijphart 1991, 1 In this dissertation I use the phrases ‘number of parties’, ‘party system size’, and ‘party system fragmentation’ interchangeably. 2 The existence of many parties typically implies the legislative presence of minority voices which can improve policy outputs and potentially bolster the legitimacy of the regime (Lijphart 1991) There is some evidence that political systems with many parties produce a close congruence between median voter preferences and government policy positions, though it remains difficult to disentangle the independent effect of multipartyism from the effect of electoral institutions (Budge and McDonald 2007; Powell and Vanberg 2000; Powell 2000). Likewise, the abundance of partisan options and vigorous electoral competition associated with multipartyism may increase voter turnout, though the evidence of in favour of this hypothesis has been mixed (Blais and Carty 1990; Blais and Dobrzynska 1998; Jackman 1987). 3 The relationship between democratic regime survival and party system size may be conditional upon the existence of a presidential regime. See: (Mainwaring 1993; Przeworski et al. 1996) 22004; Reilly 2001, 2006; Reynolds 2002; Shugart and Wattenberg 2003; Waisman and Lijphart 1996). Attempts at institutional engineering have been particularly aggressive in new democracies with potentially destabilizing ethnic divisions. Diverse societies, it is sometimes argued, are naturally prone to ethnic competition (Rabushka and Shepsle 1972). Mobilization around ethnic identities sets into place a zero-sum dynamic in which one group’s gain is perceived as another’s loss. Ambitious elites exploit group tensions for electoral gain, dividing ethnic groups into distinct partisan camps. Ethnic competition in diverse societies can not only produce a fragmented legislature but also generate incentives for extremism that threaten the integrity of both the regime and the state itself. The world abounds with cases of electoral competition precipitating ethnic violence and/or regime instability, with post-election communal violence in Kenya and Côte d'Ivoire being the most dramatic recent examples. But destabilizing ethnic competition is not only a problem in the developing world; even politicians in the seemingly consolidated democracy of Belgium recently found themselves unable to cobble together a government in the face of a polarized party system divided by ethnicity. Ethnically diverse countries have shown themselves to be particularly susceptible to the political ailments of deadlock and polarization. To head off the dangers of ethnic mobilization, some institutional engineers advocate for electoral systems that encourage inter-ethnic cooperation before an election.4 In this view, the ideal party system is one in which there are a small number of broad-based, non-ethnic parties that compete for votes across the entire nation. Scholarly advocates of this position recommend electoral systems containing an element of vote- pooling.5 Practitioners, however, are increasingly relying on the blunt instrument of 4 There are alternative means of managing the potentially destabilizing effects of diversity. Advocates of consociational democracy recommend institutions that enable cohesive ethnic parties that can effectively bargain for their group in grand coalitions (Lijphart 1977). Alternatively, Chandra finds that majoritarian systems in diverse countries with cross-cutting cleavages can encourage the fluidity of ethnic identities, thereby stabilizing the broader political system (Chandra 2005). Horowitz outlines the conditions that enable multiethnic coalitions, which can also have a stabilizing impact in a divided society, though his later work tends to endorse vote-pooling arrangements (Horowitz 2006; Horowitz 1985, 396-440) 5 “Vote-pooling” is a broad term used to encompass electoral systems that force voters from a number of societal groups to cooperate under a common party label in order to secure election. A discussion of vote- pooling in the context of ‘centripetalist’ theory can be found in (Reilly 2006). Preliminary discussion of vote-pooling appears in (Horowitz 1985, 365-369). 3ethnic party bans to accomplish similar goals (Hartmann and Kemmerzell 2010; Moroff 2010). Ethnic party bans are thought to short-circuit the causal relationship between ethnic diversity and party system outcomes. In theory, bans depoliticize ethnicity and prevent the party system from expanding out of control. Preventing the formation of ethnic parties, however, does not clear voters’ of their ethnic identities. Which raises the question: How does ethnicity impact party competition when ethnic parties have been prohibited? Despite the increasing popularity of ethnic party bans, there is little research regarding the way these solutions play out on the ground. In this dissertation I argue that ethnic diversity has an indirect effect on party system size, even when ethnic parties have been banned. Frequently overlooked in the discussion of institutional engineering in diverse political systems is the issue of rents and rent-seeking. I interpret ‘rent-seeking’ broadly to mean the use of state power to accrue benefits for a particular person or group at the expense of the broader society.6 It is well established that ethnic diversity tends to correlate with such variables as corruption and the provision of particularistic goods. This stylized fact is most often considered separately from the origins of party system size. Yet legacies of rent-seeking also have an independent impact on methods of partisan mobilization and the type of partisan options that emerge. Sensitivity to the formative impact of rents is central to understanding party system development, especially in diverse counties where the participation of ethnic 6 Under the umbrella of rent-seeking behaviour I include political patronage (the retail exchange of state resources for political support), personal corruption (the abuse of public office for individual gain), political corruption (the abuse of public office for political gain), and the various forms of influence peddling that typically result in state intrusions into the economic sphere. This definition is in line with recent scholarship on ‘political rents.’ For instance, Persson and Tabellini describe rents in the following terms: “Rents can take various forms, depending on specific economic circumstances: literally, they are salaries for public officials or financing for political parties. Less literally, one can think of various forms of corruption and waste in connection with public projects as ultimately providing rents for politicians” (Persson and Tabellini 2000, 8). Similarly, van Biezen and Kopecky use ‘party rent-seeking’ to refer to “the extent to which parties penetrate and control the state and use public offices for their own advantage, as opposed to the general public good” (van Biezen and Kopecky 2007, 240). This broad focus on political rents can be contrasted with a more narrow definition that limits rent-seeking to efforts by market actors to attain favorable government intervention to increase economic rents for a firm or sector (Buchanan 1980; Krueger 1974). My use of rents is also distinct from the broader conceptualization of ‘rentier states’ (L. Anderson 1987; Ross 2001). The rentier state literature examines the downstream political effects that occur as a result of revenue earned through the sale of natural resources, or ‘natural resource rents.’ Thus natural resource rents often (but not always) generate an expansion in political rents. In this project I do not attempt to disentangle the causal relationship between the two concepts and simply assume that the potential effect of natural resource revenue is captured in my various measures of political rents. 4parties has been prohibited. This theoretical claim is demonstrated through an empirical study of party system development in Indonesia. The puzzle of party system size in Indonesia Indonesia has a history of democratic breakdown and ethnic violence. The country’s first attempt at democracy was marked by debilitating political fragmentation, regional rebellion, and mass disillusionment with party politics (Feith 1962). The country’s second attempt at democracy began in a context of separatism and widespread communal violence (Bertrand 2004; van Klinken 2007; Wanandi 2002). Consequently, Indonesia’s institutional architecture has been deliberately designed so as to prevent ethnic competition and constrict the number of political parties.7 Legislative quotas and electoral rules have been modified to privilege large parties. Party registration laws effectively prohibit the formation of regional and ethnic political competitors.8 The over-riding goal has been to depoliticize regional and ethnic demands by channelling political activity into a small number of nationally oriented, broad-based parties. Given that there is a de-facto ban on ethnic parties, how does ethnic diversity in Indonesia impact partisan competition?9 More specifically, does the ban on ethnic parties prevent the party system from developing along the lines of ethnic cleavage? Here district-level results reveal a noticeable trend: since the inaugural election of 1999, the relative size of the party system seems to follow local levels of ethnic diversity. Compared to voters in homogenous districts, Indonesians in diverse districts are 7 ‘Ethnicity’ refers to a descent-based identity that can encompass linguistic, regional, cultural, or religious markers. In this dissertation I will use ‘ethnicity’ synonymously with the Indonesian identity category referred to as ‘suku.’ Suku is a sub-national identity category typically delineated by language and/or cultural practice. 8 It is legally possible for an ethnic party to exist in Indonesia; however, regional registration requirements severely restrict the opportunity to launch such a party. My use of ‘ban’ is consistent with Moroff, who classifies a regulatory regime as banning ethnic parties if there is a “proscriptive party ban” and/or a strict “representation requirement” that effectively prevents ethnic parties from arising (Moroff 2010, 622-624) Note: the province of Aceh is exempt from bans on regional parties. Within Aceh, regional parties are permitted to compete for sub-national offices. This concession is one piece of a broader peace agreement between the Republic of Indonesia and the Acehnese separatist movement. 9 The mark of an ethnic party is to exclusivity. Parties that implicitly or explicitly limit their voting base through appeals to ethnic identity can be considered ‘ethnic.’ Indonesia does not have parties based around suku, though it does have religiously oriented parties. In the broadest sense, then, it might be said that Indonesia has ethnic parties. Given that over 88% of Indonesians are Muslim, Indonesia’s major Muslim parties explicitly or implicitly exclude few voters. 5spreading their support across a wider array of parties. These district-level trends are one factor that contributes to national-level party system fragmentation. Across Indonesia’s three post-Suharto elections – 1999, 2004, and 2009 - the effective number of electoral parties (ENEP) at the national-level has crept up from 5.1 to 8.6 to 9.6. Despite electoral laws designed to produce a few non-ethnic parties, ethnic diversity appears to fracture the party system. What explains the variation in party system size across electoral districts? Why does party system size correlate with ethnic diversity despite an absence of ethnic parties? The latter question seems to contain its own easy answer: electoral fragmentation is caused by ethnic diversity. The parties may be outwardly non-ethnic, but this pose simply conceals the collection of local ethnic competitions that shape political behaviour and party system development. Yet a close look reveals evidence that confounds the ethnic voting story. First, the use of ethnic symbolism during campaigns is restrained. There are few obvious efforts to brand parties as the local political vehicle for a given ethnic group.10 Second, individual attitudes in diverse regions are relatively tolerant. Surveys report that ethnic chauvinism is only the norm in homogenous districts.11 Third, there is little evidence of ethnic groups voting as a block. Intra-district dynamics reveal that, in diverse electoral districts, voters from different ethnic groups actually tend to support the same parties, though they disperse their votes more widely.12 None of this supports a causal story in which the broad-based parties act as a veneer covering sub- national ethnic competitions. Yet the question remains: if it is not ethnicity producing the correlation between ethnic diversity and party system size, then what is it? I argue that, despite initial appearances, Indonesia’s party system in diverse areas is not pulled apart by clear patterns of ethnic voting. Rather, the party system is shaped by legacies of sub-national rent-seeking opportunities. Where bureaucracies are large and rampant corruption is the norm, the number of electoral parties tends to be high; on the other hand, regions with fewer rent-seeking opportunities tend to have fewer parties. While ethnic diversity does 10 For more on the use of ethnic symbolism, see Chapter 6. 11 Survey results appear in Chapter 6. 12 For more on block voting, see Chapter 7. 6correlate with and contribute to legacies of sub-national rent opportunities, careful examination reveals that it is the opportunity to manipulate sub-national state resources that shapes local electoral politics. Indonesia’s electoral engineers may have failed in their mission to keep the number of parties low, but their institutional choices have curtailed potentially destabilizing patterns of ethnic electoral competition. Theory My argument proposes that party systems are shaped by long-term legacies of local rent opportunities. Ethnic diversity contributes to rent-seeking opportunities, but the impact of diversity on party system size is indirect. I use the term ‘rent opportunities’ to refer to a politician’s opportunity to engage in rent-seeking behaviour. The opportunities to engage in rent-seeking are determined by two factors: 1) the extent of state resources available for manipulation; 2) the constraints preventing rent-seeking behaviour. Resources include public sector jobs, state loan programs, business licenses, and any other state service or financial stream that can be distributed by a politician. Constraints are shaped by both prevailing laws and societal norms regarding what constitutes abuse of state resources as well as the vigilance of state and non-state actors in punishing those whom violate norms and laws.13 Thus ‘high rent opportunity’ political systems have low constraints and extensive resources while, ‘low rent opportunity’ systems have strict constraints and minimal state intervention.14 I treat party systems as a dependent variable shaped by the independent variable of rent opportunities. In doing so, I reverse the typical presentation of the relationship that treats public sector outcomes like rents as dependent variables determined by party system size. To see how this claim challenges present theory it is necessary to briefly 13 Measurement of rent opportunities is methodologically difficult. Governments do not typically compile reliable statistics on the abuse of state resources. Data on the scope of state intervention is more readily available throughout multiple time periods. Throughout the dissertation I use civil service size as a key proxy for the scope of state intervention into the market and, by extension, a measure of rent opportunities. Where data is available, I also develop proxy variables using corruption perception indices, sub-national public service delivery scores, and central government transfer flows. The proxy measures and their relationship with ethnic diversity are examined in more detail in Chapter 3. 14 This conceptualization and operationalization of rent opportunities is deliberately similar to Chandra’s idea of a ‘patronage-democracy.’ I use ‘rents’ in place of ‘patronage’ to avoid conflation with traditional, non-state forms of ‘patronage’ and to emphasize that the relevant abuse of state resources is done for both political and personal gain. 7review literature on three established causal relationships among three intersecting variables: 1) ethnic diversity and party system size; 2) ethnic diversity and public goods provision15; 3) party system size and public goods provision. 1. Countries with more ethnic groups tend to have a higher number of effective parties (W. R. Clark and Golder 2006; Cox and Amorim Neto 1997; Geys 2006; Ordeshook and Shvetsova 1994; Stoll 2007; Vatter 2003). Underpinning the argument is an assumption that politicians in diverse polities maximize their likelihood of election by organizing parties around ethnic cleavages. Likewise, voters prefer supporting such parties for either material or psychological reasons. More ethnic groups lead to more viable partisan options. Electoral institutions modify the size of the effect; yet, all else being equal, ethnic voting fragments the party system in diverse polities. Ethnic Diversity → High # of Parties 2. Ethnic diversity reduces the provision of public goods by both state and non- state actors (Alesina, Baqir, and Easterly 1999; Easterly and Levine 1997; Habyarimana et al. 2007; Khwaja 2009; Miguel and Gugerty 2005). One line of argument proposes that ethnic diversity creates diversity in societal preferences. Members of different ethnic groups have different spending priorities. Even when these priorities converge, individuals do not want members of other ethnic groups benefitting from spending more than their own group. Because groups cannot reach a consensus on public goods provision, they choose to channel state money to particularistic forms of spending that can be directed to narrow support bases. Another stream of research focuses on collective action problems faced by non-state actors in diverse societies. Successful public goods projects require cooperative norms and mechanisms to punish free-riding. Individuals in diverse societies, however, face problems monitoring and punishing free-riders across 15 “Public goods” are defined as a good that is non-excludable and non-rivalarous. Rent-seeking, whether in the form of legal particularistic distribution or illegal corruption, typically implies the erosion and/or reduction of public goods provision. Because of an implied inverse relationship between rent-seeking and public goods, I treat the underlying causes of both phenomena as similar. While this dissertation is primarily concerned with rent-seeking, I use of the term “public goods” when the concept is the focus of a cited literature. 8ethnic boundaries. Taken together, we expect diverse polities to be marked by high degrees of particurlistic spending and low sanctions against anti-social behaviour. Ethnic Diversity → Low Public Goods Provision 3. Fragmented party systems overproduce particularistic goods and underproduce public goods (Chhibber and Nooruddin 2004; Tsebelis 2002). Different mechanisms have been forwarded that connect party system size to public policy outcomes. An electoral mechanisms starts with the assumption that parties in multi-party systems require fewer votes to gain office. Parties in systems with few competitors must build a broad appeal through promises of public goods provision; parties in systems with many competitors face less pressure to track to the centre and rely instead on particularistic promises to their distinctive voting blocs. An alternative mechanism suggests multi-party systems face distinctive problems in the legislature. A high number of parties increases the potential number of veto points. Moving policy through the legislature frequently requires that the veto-players receive payoff for their cooperation. These payoffs often take the form of particularistic goods for a party’s narrow constituency. Both the electoral and legislative mechanisms hold that parties in fragmented party systems focus on narrow delivery of particularistic goods to their own supporters at the expense of public goods provision. High # of Parties → Low Public Goods Provision At first blush these three stories seem to fit neatly together: ethnic diversity produces a high number of political parties, and a high number of political parties produce low levels of public goods provision (and high particularistic goods provision). This integrated causal story can account for the consistent finding linking ethnic diversity and public goods provision. Ethnic Diversity → High # of Parties → Low Public Goods Provision 9If the integrated story is true, creating a system that contains a small number of non- ethnic parties can not only reduce the amount of ethnic conflict in a polity, it can improve public policy outcomes. The problem with the integrated causal story is two-fold. First, the comparative party systems work does not provide direct evidence that a high number of ethnic groups produce party system fragmentation that follows ethnic lines. While theorizing takes place at the district level, existing empirical research on which the proposition is based relies on national level variables that are detached from on the ground realities. There is, in short, a mismatch between theory and empirics. The correlation between ethnic diversity and party system size appears robust and the mechanism linking the two variables is plausible, but the causal process has not been traced back to the district-level. Second, we do know that ethnic diversity correlates with low public goods provision even in the absence of democratic institutions (Alesina and La Ferrara 2004; Collier 2000). The cross-national evidence we have suggests democratic institutions may ameliorate the negative relationship between ethnic diversity and public goods provision. Thus we cannot assume the relationship between ethnic diversity and public goods provision is dependent upon party system size. Indeed, in the Indonesian case the correlation between ethnic diversity and sub-national rent opportunities was present before the transition to democracy and thus is clearly not driven by party system size. My argument modifies the causal chain slightly. Especially in countries that have recently democratized, the democratic regime inherits legacies of rent-seeking. Party systems that emerge in diverse political systems develop in a context where particularistic spending is common and the abuse of public office is the norm. Patterns of rent-seeking were shaped by ethnic diversity; however, these patterns independently affect party system development. This account places party system size at the ends of the causal chain. Ethnic Diversity → High Rent Opportunities → High # of Parties While the treatment of party system size as a dependent variable in the relationship is new, we have long known that rent opportunities impact parties as 10 individual units. Shefter illustrates how relative access to state patronage has a long-term impact on the means by which parties appealed to voters (Shefter 1977). Likewise, Chandra argues that mobilization based on identity is facilitated by access to state patronage (Chandra 2004). My argument extends these basic insights to consider the relationship between rent opportunities and party system size. Incorporating sensitivity to rent opportunities is particularly useful when analyzing party system development in countries where access to political office is one of the few lucrative areas of economic activity. Mechanism The rent opportunity theory advanced here views party system outcomes as a product of candidate goals and strategies. A high level of local rent opportunities shape candidate behaviour in three ways. First, there are higher rates of candidate entry. The state dominates economic life and attracts the attention of ambitious elites. Consequently, a high number of elites enter the political sphere looking to access local rent opportunities. Second, viable elites affiliate with minor parties. As attention shifts from national issues to local rent distribution, the political programmes of the major national parties carry less appeal to voters and elites. Elites feel free to use party labels opportunistically, seeking out parties that minimize the costs associated with affiliation. Third, voters anticipate local rent sharing and offer support to candidates – and, by extension, parties – based on a candidate’s perceived ability to deliver local resources after an election. A focus on candidates rather than party platform leads voters to support an array of minor party labels. All three factors – increased entry, opportunistic affiliation, and candidate- centered competition – combine to fragment the local party system.16 16 High levels of local rent opportunities do not always produce fragmented party systems. The relationship is conditional upon expectations of local rent sharing. If there are expectations that local rent opportunities will be monopolized by a large winning party then strong incentives exist to join the local party that can form a majority within the local legislature. Rather than using labels opportunistically, elites line up to join the local party machine. Voters anticipate control of rent opportunities by the local machine and exchange their support for promise for particularistic goods delivery. These processes generate a consolidating effect on the local party system. I return to the subject of rent sharing expectations in Chapter 2. 11 The importance of the Indonesian case The primary country of interest in this dissertation is Indonesia. Indonesia, in many ways, justifies focused study due to its demographic profile alone. With a population over 236,000,000, Indonesia is the fourth most populous country in the world and the third most populous democracy. It is also the world’s largest democracy with a primarily Muslim population. In terms of political institutions, Indonesia is the world’s most populous democracy using a proportional representation electoral system. Accordingly, Indonesian legislative elections involve the highest number of candidates in the world. Despite importance of the Indonesian case, the country’s party system is often overlooked by comparative scholars. Likewise, the evolution of the post-Suharto party system has received only minimal attention from country specialists (Choi 2010; Ufen 2008). Beyond its impressive size, Indonesia makes a compelling case for the purposes of this dissertation because it contains significant internal variation on the key independent variables of interest. First, ethnic demographics vary across the archipelago. Some provinces, like Gorontalo and Bali, are dominated by one group. Others, like North Sumatra and Bengkulu, contain citizens from across a wide array of groups, with none clearly dominant. Indonesia has a range of different sub-national ethnic structures that could, in theory, produce divergent party system results. Second, rent opportunities also vary. While Indonesia is frequently perceived as a corrupt, the opportunities to manipulate state resources are not uniform across the country. Transparency International (TI) finds considerable variation in corruption perceptions across Indonesian cities.17 Likewise, across provinces there is considerable variation in the percentage of modern sector workers employed by the government. There are meaningful regional differences in the resources available to politicians and the constraints on their behaviour. Moreover, control of most state resources has been firmly in the hands of sub-national politicians since the implementation of Indonesia’s sweeping decentralization laws passed in 1999. Variation in both key independent variables of interest make Indonesia an ideal context in 17 See: Mengukur Tingkat Korupsi di Indonesia: Indeks Persepsi Korupsi Indonesia 2008 dan Indeks Suap (Measuring the Level of Corruption in Indonesia: Indonesia 2008 Corruption Perception Index and Bribe Index). 12 which to study the causal relationship between rent opportunities, ethnic structure, and party system size. In applying the argument to Indonesia I make two basic claims. First, ethnic diversity correlates with rent opportunities and this relationship pre-exists the transition to democracy. Second, since the inaugural election voters and elites anticipate rent sharing. One unintended consequence of the new power-sharing dynamic has been the transformation of electoral politics in regions with high rent opportunities. The major parties no longer monopolize access to the state. Elites are more likely to enter electoral politics and viable candidates are more likely to join minor parties, a tendency that reflects their relatively local goals. In electoral districts with high levels of local rent opportunities, machine politics has been replaced with a candidate-centered partisan free- for-all. Electoral fragmentation is one product of this style of competition. It should come as no surprise to observers of Indonesia that a primary motivating factor for political action includes the ability to manipulate state resources. However, there are also case-specific reasons to study the origins of party system size in this case. Electoral fragmentation, when combined with the country’s bundle of electoral laws, is pushing Indonesia’s proportional electoral system to the brink of systemic failure.18 At the district level, vote wasting reaches surprising extremes.19 There are large groups of voters, especially in the Outer Islands, that lack any representation in the national capital.20 Geographic concentrations of disaffected voters can potentially challenge the 18 Electoral systems are judged by the outcomes they are meant to produce. Proportional systems are supposed to provide a relatively accurate translation of votes into seats. In concrete terms, proportional systems seek to minimize the number of ‘wasted’ votes, or votes that do not contribute to electing a legislative representative. Given that the number of seats in a district is finite, proportionality is never perfect. Nonetheless, proportional systems are supposed to limit wasted votes, at least when compared to their majoritarian counterparts. For more on evaluating electoral systems, see: (Shugart 2008) 19 In 2009, for instance, 62% of all votes in North Maluku were cast for parties that failed to win a seat in the district. This high percentage was, in part, due to a low district magnitude of 3 seats. But it is not simply the mechanical effect of seats-per-district that produces wasted votes. In East Nusa Tenggara 1, a district with 6 seats, 55% of votes were wasted. Neighbouring West Nusa Tenggara, a district with 10 seats, saw 40% of the electorate cast wasted votes. Over half of all districts saw 30% or more of the electorate cast votes for parties that do not win a seat within the district. For many of these voters, their preferred party failed to win a single seat across the entire country. In North Maluku, East, Nusa Tenggara 1, and West Nusa Tenggara the proportion of voters supporting parties with no national seats was 35%, 40%, and 34% respectively. 20 By “Outer Islands” I refer to all Indonesian islands excluding Java and Madura. This common shorthand is used to distinguish the densely populated areas most influenced by Javanese culture from the 13 legitimacy of the democratic regime. We want to know the motivation for these electoral choices and be cautious of any problems that can arise from the patterns of partisan support. The Indonesian electoral system is a work in progress and information about the successes and failures of the current set of laws can inform the institutional designers of tomorrow. Electoral fragmentation also affects sub-national governance. The Indonesian electorate tends to straight-ticket vote, meaning they support the party at all three levels of governance. Accordingly, sub-national legislatures are brimming with minor parties. This process of partisan fragmentation has coincided with the large scale transfer of fiscal and administrative responsibilities to municipal level governments. The decentralization of authority over key social services has ensured the importance of sub-national politics in post-Suharto Indonesia. While we have no comprehensive studies on the functioning of local legislatures, it is likely that sub-national party system affects the process of policy-making. It is thus helpful to understand the origins of local fragmentation. Contributions Beyond accounting for political developments in Indonesia, the dissertation makes three contributions to the theoretical literature. First, the argument is novel in that that it sets rent opportunities as an independent variable producing variation in party system size. This presentation reverses the typical formulation of the relationship, which sees governance outcomes as a product of party system size. We have long known that rent opportunities affect parties as individual units, from the way they campaign to their form of organization. I connect rent opportunities to system-level outcomes and provide a detailed causal-chain to bolster my claim. Second, the research provides insight into the complex relationship between party systems and ethnicity. The results of my study suggest party systems can be engineered so that outright ethnic competition is minimized. Nonetheless, the relationship between ethnic diversity and party system size is not dependent upon explicit or even implicit mobilization around ethnic cleavages. Ethnic diversity shapes social norms and legacies comparatively less densely populated islands that contain the bulk of the Indonesia’s landmass but only about 40% of the country’s population. 14 of state development which can have an independent impact on party systems. Institutional engineers may seek to ‘depoliticize’ ethnicity, but ethnic diversity will continue to have long lasting unforeseen effects on political development. Third, I extend several of the insights from the literature of party nationalization to account for district-level outcomes. Major works on party system nationalization begin their theorizing with the incentives candidates face to coordinate their actions in either national or local parties (Aldrich 1995; Chhibber and Kollman 2004; Hicken and Stoll 2008; Hicken 2009). This literature underlines how the balance of local verses national power shapes candidate affiliation strategies. When candidates have national goals they coordinate in national parties; when candidates have local goals, they coordinate in local parties. All of the work, however, focuses on explaining the size of the national party system. I argue that the localization of politics can impact candidate coordination both across and within electoral districts. When candidates organize to capture local prizes there can consequences for district-level party system outcomes. Plan of dissertation This dissertation is structured around the premise that party system outcomes reflect strategic choices made by individual candidates. Many of the key choices that determine the eventual number of parties are made before the electorate even casts a ballot. A careful analysis of electoral results is, of course, an essential component in the study of party systems. To fully understand outcomes, however, we must look closely at pre- electoral candidate decision-making: why candidates enter politics, how they organize themselves into parties, and how they win support from the public. Rather than study the Indonesian party system by region or temporal period, I structure empirical chapters around distinct aspects of the candidate experience: entry, affiliation, and campaign. These chapters precede the final exploration of party system outcomes. The chapters will proceed as follows. Chapter 2 lays out my theory of party system size. First I explain why existing theories of electoral institutions and ethnic voting cannot fully explain how party system correlates ethnic diversity in the Indonesian case. I then present a theory of party system size in which local rent opportunities and power sharing determine the number of parties in the system. Chapter 3 presents an 15 overview of the institutional, demographic, and partisan context of the Indonesian case. I also demonstrate that rent opportunities tend to correlate with ethnic diversity across Indonesia’s sub-national units. The next three chapters examine distinct links in the causal chain. Chapter 4 asks: why do we see variation in cross-district rates of candidate entry? While entry rates do correlate with ethnic diversity, I show that access to rent opportunities is actually motivating these entry decisions. Chapter 5 asks: why do viable elites affiliate with minor parties? To answer this question I examine patterns of career building and party switching through time. I find that candidates in provinces with high rent opportunities tend to switch parties, demonstrating that party identification is valued less by elites and voters in these areas. Chapter 6 asks: under what conditions do voters choose parties based on connections with a candidate? Through a close analysis of the 2004 election, I find that the proportion of voters casting an optional ‘preference vote’ closely correlates with rent opportunities. The chapter demonstrates the personalist nature of electoral competition in those areas where candidates can credibly pledge to provide post-election goods to supporters. All three chapters together support my argument that, in high rent regions, more aspiring politicians are more likely to enter politics, more likely to use party labels opportunistically, and more likely to run a campaign based on delivery of particularistic goods. Chapter 7 turns to party system outcomes. Through an examination of intra- district dynamics, I parse out the independent effects of ethnic diversity and rent opportunities on party system size. I demonstrate that rent opportunities have a significant effect on electoral behaviour even in the absence of ethnic diversity. Chapter 8 shifts from a static to a dynamic analysis of party system evolution in Indonesia. In this chapter I take on the question: why have some electoral districts experienced more electoral fragmentation than others? I link the expansion of district level party systems to rent opportunities. Specifically, I show that the decline of Golkar’s party machine in high rent areas precipitated the extreme fragmentation of local party systems. Finally, Chapter 8 offers conclusions and implications for the study of comparative party systems. 16 Chapter 2 – Theory Why do some districts have many parties and others only a few? I argue that local rent opportunities influence party system size. The extent of local rent opportunities can vary across a country and are structured, in part, by a region’s ethnic demography. Local rent opportunities produce patterns of elite political entry and party affiliation that can lead to severe district-level electoral fragmentation. By studying variations in party system size across electoral districts we see party systems being shaped by long-standing patterns of rent-seeking. This helps provide a more accurate knowledge of the electoral environment Indonesia’s engineers are trying to manipulate. The chapter proceeds as follows. The first section defines the key dependent variable of interest – district-level party system size – and demonstrates the considerable variation that exists within Indonesia. In the second section I review the current literature on the determinants of party system size, explaining why none of the existing theories can account for variation in Indonesia while emphasizing the useful insights offered by each. The third section presents a theory linking party systems outcomes to local rent opportunities. In this section I define my use of rents and explain why local rent opportunities shape patterns of candidate entry, party affiliation, and campaign strategies. I go on to link these patterns of party organization and rent opportunities to electoral outcomes that determine party system size. The fourth section discusses the project’s broader contribution to the field of comparative politics. Electoral fragmentation in Indonesia Indonesia uses a system of proportional representation (PR), but like most PR systems it sub-divides the country into separate electoral districts. Parties construct a party list for each electoral district and voters face a distinct slate of candidates from district to district. Since 1999 there has been an expansion in the number of districts and a reduction of the average district magnitude. 21 These reforms reflect efforts to bring politicians closer to their constituents and reduce the number of parties in the system. 21 ‘District magnitude’ refers to the number of seats allocated to an electoral district. See: (Cox 1997; Reed 1990; Taagepera and Shugart 1989) 17 The effective number of electoral parties in an electoral district has varied widely.22 Since the fall of Suharto there have been three open elections for the national legislature. In the inaugural election of 1999 there was an average of 4.3 electoral parties per district. 1999 would turn out to be the election with the lowest number of parties, however. The election of 2004 witnessed a significant increase in the number of parties in the district, with the average reaching 7.1. The considerable variation was further evidenced by the spread between the district with the most concentrated vote (3.2) and the district with the most fragmented vote (11.5). In 2009 the mean number of parties per district crept up to 8.5. Table 1 – Party system size in national electoral districts: descriptive statistics Year Number of Districts Mean Magnitude Mean ENEP Standard Deviation Minimum Maximum 1999 27 17.1 4.3 1.3 1.6 6.6 2004 69 8.0 7.1 2.0 3.2 11.5 2009 77 7.3 8.5 1.9 4.3 13.6 There has been considerable variation in the number of electoral parties across Indonesian electoral districts. Existing comparative party system theories, however, cannot account for the outcomes in the Indonesian case. Explaining electoral fragmentation: state of the literature The study of party system size has produced a rich and active literature. I will review four approaches that account for electoral fragmentation: 1) institutional; 2) sociological; 3) interactive synthesis; 4) aggregative. Each approach either emphasizes the importance of a distinct independent variable or blends the variables in novel ways. For each approach I will explain why our current theorizing cannot account for the Indonesian case while highlighting the useful insights that I will build off when developing my theoretical model. 22 In this chapter the ‘effective number of electoral parties,’ ‘electoral fragmentation’, and the ‘size of the party system’ are used synonymously. The effective number of electoral parties is derived from the inverse of Rae’s fractionalization index: The standard mathematical expression for the effective number of electoral parties is: ENEP = 1/(si) 2where si is the proportion of votes for party i. 18 Electoral institutions Maurice Duverger is credited for making the theoretical link between party fragmentation and electoral institutions. The oft-cited Duverger’s Law asserts that plurality systems in single-member districts tend to produce two-party competition. The addendum to Duverger’s sociological law proposes that proportional systems tend to produce multi- partyism (Riker 1982). Whereas Duverger refers to allocation rules, contemporary scholars have applied Duverger’s original insight to district magnitude. Reed (1990) argues that multi-member districts tend to produce competition among n + 1 parties (where ‘n’ equals the district magnitude). Cox asserts that M + 1 (where ‘M’ equals district magnitude), “imposes an upper bound on the effective number of competitors that will appear in equilibrium” (1997, 139). If political actors are acting strategically the number of effective parties should not rise above the upper bound. In sum, the basic logic of the institutionalist literature is that an increase in district magnitude causes an increase in the fragmentation of the party system. Before examining the institutionalist argument in the Indonesian context it is important to underline the conditions required for the M+1 rule to take effect. There are two assumptions about the elite behaviour as it relates to elite entry. First, there must be common knowledge of electoral frontrunners. If elites are to avoid unwinnable races they need to know where they stand vis-à-vis their competitors, and as such they need to hold accurate information about local races. Second, elites must be concerned with the immediate electoral competition. If the goal of the campaign is not necessarily winning but the promotion of the candidate or some other cause then the field of candidates will not necessarily narrow. When elites are focused on the current election and have accurate information, marginal political contenders should not enter the race. This winnowing of the electoral field reduces the fragmentation of the vote. Voters, too, act strategically in ways that reduce the fragmentation of the vote. There are three conditions under which voters can be expected to act strategically: 1) they are short term instrumentally rational; 2) they have access to publicly available, reasonably accurate information of candidate strength; 3) they adjust they vote based on new information. If voters are not focused on the immediate contest, if they do not have access to accurate information, and/or they do not adjust their vote intensions with new 19 information, then those candidates with little chance of winning will not be abandoned and the electoral vote will be fragmented. The theoretical conditions required for M+1 to hold are, at best, only partially present in the Indonesian case. Elites and voters have minimal district-level information. Previous electoral results, especially incumbency, provide a snapshot of local strength. Given the small number of free elections the country has experienced, however, the accuracy of this information is limited. Of more relevance to this dissertation, however, is the complex structure of incentives that lead voters to support marginal candidacies. Voters regularly receive material support, through either direct gifts or the provision of community goods, which weigh on electoral decisions. A citizen who uses her vote to extract immediate gifts from potential patrons is only marginally concerned with the outcome of the race. Indonesia’s simultaneous multi-level elections add a further level of complication. As I will expand on below, both elites and voters can tilt their attention to either national or sub-national office.23 Even if voters and elites are concerned with supporting winners in the immediate electoral competition, it is not necessarily the national level competition that captures their attention and motivates their strategic decisions. Empirically, there is no evidence to suggest that Indonesian candidates either abandon hopeless campaigns or that voters abandon hopeless candidates due to the pressures of the electoral system. Figure 1 shows there is no correlation between district magnitude and the number of electoral parties in a district.24 Elites that run for national office have goals that go beyond simply winning office and thus have little incentive to strategically exit the race. Also, the multi-level nature of Indonesian elections complicates the voters’ task of calculating party strength from publically available information, making strategic voting less likely. On its own, the electoral system cannot explain variation across districts. 23 The ideas that spill-over exists between different levels of electoral competition is also present in the extensive literature on presidential ‘coat-tails.’ This literature explores the effect of executive competition of legislative elections. See: (T. D. Clark and Wittrock 2005; W. R. Clark and Golder 2006; Cox and Amorim Neto 1997; Filippov, Ordeshook, and Shvetsova 1999; Golder 2006; Samuels 2002; Shugart 1995) 24 Note: results from the 1999 election do not appear on Figure 1. Because district magnitude varied widely in 1999, the inclusion of this year distorts the appearance of the scatter-plot. No discernible trend existed between magnitude and the effective number of electoral parties in 1999. 20 Figure 1 - Electoral parties and district magnitude Social cleavages Comparative party system scholars have a long tradition of tracing party system outcomes back to a country’s socio-economic and cultural cleavages (Lipset and Rokkan 1967; Lipson 1964; McRae 1974; Meisel 1974; Rose and Urwin 1969). This body of literature hangs loosely around the idea that voting patterns reflect deeply entrenched social divisions. Countries have complex histories that generate certain societal cleavages and the party system reflects these social realities. Electoral systems, rather than being determinative of party systems, are determined by the country’s political forces. Thus the presence of either configurations such as PR and SMDP are a consequence of an emerging party system rather than a cause of a certain party system. The classic formulation of this argument is Lipset and Rokkan’s investigation of cleavage structures in the Western democracies. The authors point out that modern party systems tend to reflect the social divisions that animated party competition around the time of full manhood suffrage. These systems involved parties that reflected the primary cleavages that arose in modern states.25 More important to the freezing hypothesis is that 25 The first set involved two cleavages that arose during the national revolution: 1) the conflict between ecclesiastical and state authorities over the rights and place of the church; 2) the conflict between the administrative centre and the periphery, which encompassed the conflict between a dominant and sub- ordinate ethnic group. The second set of conflicts involved two cleavages that arose during the industrial revolution: 1) the conflict between agricultural and commercial concerns; 2) the conflict between workers Effective Number of Electoral Parties by District Magnitude [2004, 2009] R2 = 0.0295 0 2 4 6 8 10 12 14 2 4 6 8 10 12 District Magnitude Eff ec tiv e N um be r o f Ele cto ral Pa rtie s 21 early mobilization tends to stick. Parties built organizations that bounded their memberships and voting blocs, a process that strengthened loyalties and reinforced the existing pattern of competition. The sociological approach has several relevant limitations. First, more cleavages do not necessarily mean more parties. Not all cleavages are ‘partisised’ in all countries, and the very presence of multiple cross-cutting cleavages may mitigate against party formation (Meisel 1974, 6). Second, the approach is focused on explaining outcomes at the national level. In Lipset & Rokkan’s study it is national policy battles – disestablishment, language of education, tariffs rates, extent of redistribution, etc – that drive party formation. How can we use this framework to explain district contests? The most straight-forward way would be to attempt to gauge the relative strength of the different groups within a district. This effort would have to be sensitive to which cleavages are mobilized nationally in order to properly measure the district’s heterogeneity in terms of mobilized groups. The complex nature of the issue places limits on both the theorizing and systematic study of district-level relationships between cleavages and party systems. The most common application of the social cleavage approach to Indonesia traces the roots of Indonesia’s contemporary party system back to societal divisions of aliran.26 In 1955 the four major parties collected 90% of the vote and each party was linked to a particular aliran group. As many have pointed out, however, today the connections between aliran and the party system are tenuous (Liddle and Mujani 2007; Ufen 2008). Even if we assume that partisan support circulates within boundaries set by aliran, each group now has an array of parties to choose from and some parties are clearly able to transcend these divisions. This complicates the process of mapping aliran divisions onto party system outcomes. For instance, the Javanese areas that were thought to have the deepest and most varied aliran divisions in 1955 now report some of the more consolidated party systems. While the debates about aliran continue to animate scholarly and producers. Which cleavages were mobilized around the time of suffrage expansion was influenced by a range of particularistic factors. 26 Aliran is a term that refers to broad socio-cultural groups differentiated primarily by religious practice and, to a lesser extent, class. The relationship between aliran and the Indonesian party system is taken up in more detail in Chapter 3. 22 discussion of the Indonesian party system, the concept provides minimal leverage for those interested in the variation of party system size across electoral districts. The interactive synthesis The rough division between those attempting to explain party system size in terms of either institutions or sociological factors has eroded in recent decades. Contemporary scholarly efforts offer a synthesis of these two views, arguing that party system size is an interaction between a country’s social diversity and its institutional structure. There is a lively discussion about which institutions determine outcomes, how core concepts should be measured, and how statistical models should be constructed. Nonetheless, the basic understanding has been set: Number of Parties = Institutions X Social Cleavages. Numerous cross-national studies lend support for this formulation (W. R. Clark and Golder 2006; Cox and Amorim Neto 1997; Ordeshook and Shvetsova 1994; Stoll 2007). The interactive approach has a straight-forward mechanism to explain why social diversity tends to fragment the party system. In this story, social cleavages create a number of social groups. Individuals within these groups are conscious of the group identity. These groups may or may not mobilize themselves politically by forming (and voting for) a political party that represents their group interests. Political parties, then, are rooted in social groups. However, the ability of a social group to mobilize behind a party is conditioned by the existing electoral system. Where an electoral system has a low effective threshold for party entry (e.g. high district magnitude, no national threshold requirements), social groups should mobilize behind a party that claims to represent the group’s political aspirations. Operationalizing the number of social cleavages and mobilizable groups within a country is a challenge. In the recent cross-national studies, ethnic diversity has become the primary social cleavage variable used to measure the number of groups, supplanting all others. According to this measure, more ethnic diversity means more groups. This raises several problems. First is the question of whether ethnicity should be treated as the dominant cleavage. It is taken as given that each country has a number of ethnic groups and under certain institutional conditions these groups will mobilize politically by forming a partisan vehicle to represent their interests. Yet those working using a 23 sociological approach have been mixed opinions on the primacy of the ethnic divide. The underlying political primacy of the ethnic cleavage remains an open question. 27 Second, the approach is unclear how the district should be treated. Theoretically, the electoral system variables all affect district-level competition and they are constructed in ways that reflect this. Ethnicity, on the other hand, is measured at the national-level using aggregate demographic data, meaning the variable captures no information about district-level ethnic composition. In part this reflects data constraints. Yet not all ethnic groups will form political parties; the relative size and dispersion of an ethnic group will affect its likelihood of mobilizing around a political party (Chandra 2004). It is, for instance, unlikely that small groups isolated to only a few districts will form a political party. Because district-level data do not tell us about national size and dispersion, it is unclear whether we should expect the same effects when studying district dynamics.28 Third, the interactive approach has an implication that is both overlooked and relevant to the study of Indonesia; namely, the predicted non-effect of ethnic diversity in countries that prevent ethnic party formation. The causal line between latent ethnic diversity and party fragmentation requires that a group (ethnic or otherwise) has the ability to form its own political vehicle. Where entry is barred or otherwise prevented, as it is in Indonesia, the causal story falls apart and the empirical predictions are unclear. Despite the non-existence of ethnic parties, we do see a simple correlation between the number of electoral parties and the number of ethnic groups within a district.29 Figure 2 plots the effective number of electoral parties by ethnic fractionalization. In the elections since 1999, it has been the most ethnically homogenous 27 Lipson, for example, takes ‘race’ as the most important ascriptive cleavage because it is neither chosen nor easily hidden (Lipson 1964). However, a study of developed world party systems led Rose and Urwin to the conclusion that ethnic groups tend not to form political parties and suggest Lipset & Rokkan may have overemphasized the importance of this cleavage (Rose and Urwin 1969). Rose and Urwin’s findings are likely skewed by the universe of cases populated primarily by European democracies. The flexibility of borders in the European experience often allowed peripheral ethnic groups to use their ‘exit’ option of secession rather than the ‘voice’ option of political party mobilization. Systematic studies of the ‘ethnification’ of party systems across countries in both the developed and developing world have only recently begun. See: (Chandra and Wilkinson 2008) 28 As Taagepera points out, the discipline does not yet have a simple index of ethnic geographical concentration that captures the intuition that the dispersion of groups has an effect on ethnic mobilization (Taagepera 2007, 278-279). 29 The standard mathematical expression for the level of ethnic fractionalization is: Ethnic Fractionalization = 1 - (si) 2where si is the proportion of the population in ethnic group i.. 24 electoral districts that produce the fewest number of electoral parties.30 For example, in 2004 the three districts with the least number of electoral parties were Bali, Gorontalo, and Central Java IV, all three of which were dominated by members of one ethnic group. In 2009 the list included Bali, this time with Aceh II and West Java I. Again, all three are relatively homogenous districts. This preliminary consideration does not take into account the interactive aspect of the argument but it is suggestive that some dynamic between party systems and ethnicity is at play in Indonesia. Figure 2 - Electoral parties and ethnic fractionalization Despite the suggestive correlation, problems exist with the causal mechanism. In Indonesia there is little evidence that ethnic diversity produces a high number of politicized groups that can be mobilized by elites. According to opinion polls, voters in diverse electoral districts do not exhibit strong in-group preferences. Inter-ethnic political tolerance is also exhibited at the candidate level, where campaigns are restrained in their use of ethnic symbolism and ethnic appeals. At the district level, ethnic groups rarely vote as a defined block. Diverse districts have more parties, but there is no sign that the mechanisms suggested by the interactive approach are playing out in the Indonesian case. The empirical puzzle offers an opportunity to advance theoretical knowledge as to how ethnic diversity has an effect on party systems. To account for this empirical puzzle I 30 There is no discernible relationship between ethnic fractionalization and the effective number of electoral parties in 1999. I will return to the deviant 1999 election results in more detail throughout the dissertation. Effective Number of Electoral Parties by Ethnic Fractionalization [2004, 2009] R2 = 0.1637 0 2 4 6 8 10 12 14 0 0.2 0.4 0.6 0.8 1 Ethnic Fractionalization [0-1] Eff ec tiv e N um be r o f Ele cto ral Pa rtie s 25 develop an alternative theoretical story that can explain party system outcomes in Indonesia and, potentially, across the developing world. The theoretical building blocks of my argument are found in an alternative account of party system size I refer to as the party aggregation approach. Party aggregation There is a growing party systems literature that links the number of electoral parties to the concentration of policy-making authority within the political system. Elites organize their party-building efforts and voters modify their electoral behaviour depending on where authority is located within the system. The centralization of authority provides incentives for party aggregation, defined as the extent to which electoral candidates coordinate their efforts across districts under one party banner (Hicken 2009, 2). Where the institutional incentives for party aggregation are low, the effective number of parties within the system multiplies. The key insight of the approach is to focus on the strategic calculations of legislators, candidates, and voters. Legislators face coordination problems and must form organizations in order to effectively accomplish individual and collective goals. The organizations that form offer recognizable party labels that communicate information about policy programs and past performance. Voters can use the information from party labels to assist them in voting for candidates that can pursue their policy preferences. Office seeking candidates want to win elections and access power. Accessing power in the legislature requires working in a party and winning election requires successfully appealing to the electorate. This need for parties forces candidates to coordinate their activities across electoral districts. When facing the choice of what type of party to affiliate with a candidate considers the relative centralization of policy-making authority. There are two basic forms of centralization: vertical and horizontal (Hicken 2009). Vertical centralization refers to the relative policy-making power of the national government over the sub- national units. Horizontal centralization refers to the relative policy making authority of a given branch of government. In the archetypical centralized polity, policy-making authority is concentrated in a national level cabinet dominated by one party. 26 Candidate goals and party affiliation strategies are determined by the dispersion of power within the system. If, for example, there is a strong national executive then party labels that dominate the executive competition will draw the most voter attention. Candidates in such a centralized political system will face pressure to associate with a party that can compete for this prize (Chhibber and Kollman 1998, 2004; Hicken and Stoll 2008; Hicken 2009). The same is true for vertical centralization. When authority lies in the national capital, candidates gain the most from joining parties with national policy goals that can play an important role in the national legislature. As such, there are strong incentives for candidates to aggregate across electoral districts and join national parties. Attaining national power is less important in countries that devolve significant budgetary authority to sub-national units. A space opens for the formation of regionally focused parties. In these circumstances, even a candidate running for national office may find it beneficial to associate herself with a regional party as this label communicates to the voters more pertinent message than a national party label. While the aggregative approach has been used to explain national party system outcomes across nations and within the same nation over time, electoral institutions are the main independent variable used to account for the number of electoral parties within the district. Because the typical variables of interest tend to encompass the entire country (e.g. authority of the cabinet, authority of the provinces) they cannot easily account for the number of electoral parties across districts. National electoral districts in Indonesia are typically made up of multiple municipal administrative units that retain substantial policy-making authority.31 A simple extension of the aggregation approach would suggest district level fragmentation is due to low cross-municipal coordination. Plotting the electoral fragmentation of national electoral districts by the average fragmentation of the municipalities contained within those districts provides a quick test of the hypothesis. A substantial difference between 31 Contemporary Indonesia has three levels of constitutionally protected governance: national, provincial, and sub-provincial. Sub-provincial units are referred to as kabupaten and kota. Kabupaten tend to be larger units roughly equivalent to the North American ‘county,’ while kota tend to include only the boundaries of one city. Collectively, Indonesia’s sub-provincial units are most often referred to as ‘districts’ in English. Given that this dissertation is expressly concerned with electoral districts, I will attempt to avoid confusion by referring to sub-provincial units as ‘municipalities.’ 27 the two variables would indicate the presence of party system inflation.32 As demonstrated in Figure 3, however, there is a tight correlation between the two variables. Party system size in national electoral districts is clearly driven by electoral fragmentation within municipal units rather than a failure of coordination across municipal units. Figure 3 – Party system aggregation The lack of party system inflation within electoral districts has one additional implication: it strongly indicates that ethnic groups do not vote a group. Within a diverse electoral district, different ethnic groups dominate different municipalities.33 Thus if co- ethnics were to vote as a group, we would expect low to modest levels of party system fragmentation at the municipal level and a sharp difference between the municipal level averages and the electoral fragmentation of the national district. At least in 2004, this was 32 Technically, measures of party system inflation calculate the difference between average district level outcomes and broader national/provincial level outcomes. For present purposes, however, a simple comparison of the two values will suffice. The issue of inflation and intra-district aggregation is examined more thoroughly in Appendix B. No evidence exists to suggest district-level party system size is caused by the failure of political elites to coordinate across municipalities. 33 Take the electoral district of East Nusa Tenggara II for instance: Kabupaten Sumba Barat is dominated by suku Sumba, Kabupaten Timor Tengah Selatan is dominated by suku Atoni Metto, and Kabupaten Belu is dominated by suku Belu. This type of municipally bounded ethnic segregation is common. 28 clearly not the case.34 The non-importance of inflation strongly suggests that cross- district variation in party system size is not caused by distinct patterns of communal voting in diverse districts. In sum: The four existing approaches that provide a causal explanation for party system size cannot account for cross-district party system size in Indonesia. District magnitude clearly does not explain cross district variation. There is evidence to support the sociological/interactionist argument that the internal ethnic structure of the district can potentially explain party system size; nonetheless, the mechanism is complicated by the non-ethnic nature of the Indonesian party system. Though district level results are not caused by intra-district failure to aggregate across municipalities, the approach does draw our attention to the ways in which sub-national governments affect political organization. Struggles for control of local resources can shape the party affiliation decisions of elites competing at the national level. I build off this base to construct a theoretical approach that can account for outcomes in Indonesia. A theory of party system size across electoral districts Local rent opportunities play a decisive role in shaping electoral outcomes. In high rent districts, the benefits for attaining local power are a driving force behind party organizational efforts.35 The elevation of local issues (rent access) creates space for the success of minor parties. Local rent opportunities induce high levels of candidate entry, entice viable elites to affiliate with minor parties, and incentivise personalist campaigns. These factors increase the number of viable options facing voters. In short, the struggles for local resources explain variation in party system size across Indonesian districts.36 The argument raises a distinct set of conceptual questions that must be answered: What are rent opportunities and why do they vary? Why do rent opportunities have an effect on candidate entry, party affiliation, and the personalism of campaigns? How do 34 Even less intra-district party system inflation existed in 1999. Though I do not have the data to systematically investigate party system inflation in the most recent election, preliminary observations suggest the 2004 pattern of low intra-district inflation occurred in 2009 as well. 35 ‘High rent district’ refers to a national electoral district that contains sub-national governments with high levels of rent opportunities. 36 The two levels of elected sub-national government that allow partisan organization in Indonesia are the provincial and the municipal. All references to ‘sub-national’ and ‘local’ government will refer to both municipal and provincial governments. 29 these candidate level factors affect party system size? Under what conditions do we expect the causal mechanism to occur? Defining rent opportunities Rent opportunities are determined by two factors: 1) state resources available for manipulation; 2) constraints against rent-seeking behaviour. Constraints are low where law enforcement officials are lax and societal norms encourage politicians to use state resources to support their political networks. Resources are high when politicians have a large budget to skim from, plenty of government services to pass out as favours, and a large number of jobs to distribute to supporters. A combination of low constraints and extensive state resources make for high rent opportunities. But why would rent opportunities ever vary across a country? One existing answer ties rent opportunities to sub-national ethnic structures. Ethnic diversity has an effect on both state resources and constraints on behaviour. Regarding resources, ethnic diversity can influence budgetary and spending decisions, thereby expanding resources available for politicians to distribute. Attempts at modelling budgetary decision-making in diverse societies typically assume divergence of tastes which can come in two forms (Alesina, Baqir, and Easterly 1999). First, individuals in different ethnic groups may prefer spending on different non-excludable projects (e.g. hospitals vs. libraries). Conflict over priorities leads to under-supply of non-excludable goods. Second, citizens may agree on priorities but only value spending if members if co-ethnics gain more than other groups. Thus citizens in Ethnic Group A only support hospital spending if co-ethnics are seen to benefit from such spending more than citizens from Group B. This form of ‘taste for discrimination’ leads to an under-supply of non-excludable goods, especially in situations with high between-group income inequality (Baldwin and Huber 2010). Particularistic spending, on the other hand, is more likely in diverse contexts. Inter-group log-rolls on particularistic spending solve the inherent problem of dividing the budgetary pie amongst the different ethnic groups. In more authoritarian contexts, where one ethnic- group rules, policy-makers can use the savings from the under-supply of non-excludable goods to provide particularistic goods to the ruling ethnic group. Over the long term, privileging particularistic spending has a self-reinforcing logic. Particularistic projects and agencies are created, funded, and become interest groups in their own right. Shifting 30 spending from particularistic to non-excludable goods threatens entrenched interests tied to specific groups and can disturb social stability. As such, the bias toward particularistic spending in diverse political systems is sticky. Ethnic diversity can also affect the constraints placed on policy-makers. In particular, ethnic diversity tends to correlate with various measures of corruption (Glaeser and Saks 2006; Mauro 1995). This suggests that politicians in diverse areas perceive the illegal manipulation of state funds and policy implementation as broadly acceptable. One reason corruption flourishes in diverse contexts is due to the difficulties of social sanctioning in diverse contexts (Habyarimana et al. 2007). For instance, the provision of collective goods in diverse contexts is especially prone to the problem of free-riding as those that shirk do not expect effective sanctioning across ethnic lines (Miguel and Gugerty 2005). If ethnic barriers increase the costs of sanctioning, those with clean government preferences may be dissuaded from pursuing corrupt officials. Additionally, ethnic diversity leads to lower levels of social capital and social trust (Putnam 2007). Citizens in diverse contexts may not trust their co-citizens to apply sanctions against corrupt officials and thus do not bother applying sanctions themselves. Whatever the mechanism at work, politicians in diverse political systems behave as if there are fewer constraints on their behaviour than politicians in homogenous systems. Diverse societies are associated with certain political pathologies (low trust, zero- sum competitive dynamics, etc) that increase rent opportunities in the long term. Yet an equally important part of the story connecting diversity and rent opportunities is the relationship between diversity and state structure. Social demographics shape the internal organization of the state itself. The presence of multiple groups is frequently associated with a demand for some form of group autonomy. The response of the state can impact internal boundaries, the centralization of fiscal authority, and the distribution of state revenue. These structural accommodations directly affect the local opportunities available to politicians. For instance, evidence from the cross-national literature suggests diverse countries are more likely to decentralize fiscal authority (Panizza 1999). In Indonesia, structural accommodations have resulted in comparatively high per-capita transfers flowing to the sub-national units in diverse areas. These transfer flows expand the resources available to politicians in certain areas. 31 Taken together, ethnic diversity can impact both the resources available to politicians and the constraints placed on their behaviour. While some of the proposed mechanisms suggest electoral motivations, these are not necessary to produce the relationship between diversity and rent opportunities. The thrust of this dissertation, however, is not to explain why ethnic diversity and rent opportunities might correlate. Rather, I argue that, under certain institutional and partisan circumstances, rent opportunities have an independent effect on political behaviour that produces party system fragmentation. Rents and candidates strategies: entry, affiliation, and campaigns The theory advanced here views party system outcomes as a product of candidate goals and strategies. Candidates coordinate their political activities with other candidates to achieve certain goals. Broadly, candidates may orient their goals toward the capture of either national or sub-national power. To account for party system outcomes we must understand how the prize motivating candidate behaviour can vary systematically across a country. Figure 4 - Theory mapping I argue that rent opportunities have three interrelated effects on candidate behaviour. The theoretical model appears in Figure 4. First, rent opportunities increase candidate entry. Second, rent opportunities increase the likelihood that viable candidates will affiliate with minor parties. Third, rent opportunities increase the personalist and clientelistic nature of campaigns. The combined impact of all three factors has an effect on party system size. When a high number of viable elites run candidate-centered campaigns under minor party labels, the electorate tends to disperse its vote across a wide 32 array of minor parties. Consequently, the effective number of parties in the system increases. The argument hinges on three assumptions. First, the saliency of local office access dominates other issue areas when rent opportunities are high. Second, voters and elites hold ex ante expectations that sub-national rent opportunities will be shared after the election. Though this does not require rents to be shared equally, it does imply even candidates from minor parties expect to access local resources if elected. Third, even non-elected candidates earn influence within their local party branch for participating in a campaign. This holds true for both national and sub-national candidates. Providing some co-partisans are elected to sub-national office, influence with a local branch can be used to access rents between elections. To dig into the logic of the argument I examine each piece of the causal chain – entry, affiliation, and campaigns – independently. Candidate Entry The number of viable partisan options faced by voters is, in part, dependent on patterns of candidate entry. A party with no local champions is unlikely to receive support from the populace; a party with a full slate can expect a more extensive campaign from its local agents. Accounting for candidate entry, then, provides insight as to why some voters face different numbers of viable partisan options across electoral districts. Why is there variation in candidate entry across electoral districts? While a robust formal literature exists on candidate entry, there have been few attempts to understand the phenomenon in developing political systems where policy positioning is less important than accessing state rents. Samuels (2003, 15) offers a basic starting point here. In his simple formalization, a candidate’s utility from running equals the benefit from office times the probability of winning, minus the costs of running for the office.37 This is a widely applicable model that focuses our attention on three important factors: the probability of winning, the value of the office, and the costs that will be incurred. 37 The relationship is expressed as: Ui (Running for Office o) = PioBio – Ciowhere Pio is the probability of individual “i" winning office “o”, Bio is the benefit individual “i" receivesfrom holding office “o”, and Cio are the costs individual “i" incurs for seeking office “o.” 33 Why do some districts attract more viable politicians than others? Samuels can provide a framework for answering this question but does not explain why office benefits and/or costs may vary across a country. Chandra ties entry the benefits of office to the access it provides to patronage (Chandra 2004, 2007). Chandra states: In a patronage democracy, obtaining control of the state is the principal means of obtaining both a better livelihood and higher status. Elected office or government jobs, rather than the private sector, become the principal source of employment. And because individuals who control the state are in a position of power over the lives of others, it also brings with it higher status. Those who have the capital to launch a political career in patronage-democracies, therefore, seek political office. (Chandra 2007, 87) In short, it is the ability to manipulate state resources that draws competitors into the system. Chandra provides leverage in explaining why some offices are more valuable than others. In Chandra’s account, however, the benefits of office are fixed. All polities have the same office benefits if they reach the patronage-democracy threshold. For the purpose of within country study, however, the benefits of office are better understood as more of a sliding scale. Some polities will offer more opportunities for rent-seeking than others. Rent-seeking opportunities, meanwhile, increase the value of holding office. Where expected office benefits are high there is more incentive to skimp, save, and borrow in order to invest in a political career. Entrance costs may also be large, but so is the value of office. Aspiring politicians in areas where office benefits are large are more likely to become candidates. As a result, more politicians enter the political competition. Both sub-national and national candidates are more willing to enter politics and become a candidate when office benefits are high. For the sub-national candidate, the direct benefit from holding local office tends to be correlated with the local rent opportunities. For the national candidate, the link to sub-national co-partisans through the mechanism of party influence connects their payoff to the value of local office. At both levels of government local rent opportunities entice higher levels of candidate entry. Candidate affiliation Minor parties are able to succeed when they are able to attract viable candidates. But why do politicians join minor parties? Hicken (2009) presents a simple formalization of the incentives facing national legislative candidates. Candidates are motivated by the rewards 34 for entering either a major or minor party. They calculate the expected utility of coordination for each option. The payoff is the benefits of aggregation (the size of the prize for being in the largest national party) and the probability that their efforts will allow them to enjoy this prize. These are offset by potential costs incurred for coordination. The larger the expected utility for coordinating efforts in a large party as compared to expected utility of coordinating in a small party, the greater the incentives for aggregation.38 The relative centralization or decentralization of power tends not to vary across electoral districts in a single country. Local rent opportunities, on the other hand, do vary across districts. Rent opportunities affect the size of the national prize. Where the local government plays a dominant role in the economic lives of the citizens, and personal relationships with local politicians can help access state resources, elite and voter attention tends to shift from national to local issues. Where opportunities for local rents are high the expected payoff to a candidate for coordinating her efforts in a large national party is low. On the other hand, the expected payoff from attaining influence with a local party branch is high. Minor parties can be effective if voters and elites are focused on local rents. Minor party labels typically do not provide the electoral brand recognition of the major parties. Likewise, the career trajectory of a national level politician affiliated with a minor party may be stunted by the small size of the party’s legislative caucus. On the other hand, candidates that affiliate with minor parties can often avoid the financial and ideological costs that come with participation in a major party. Moreover, if the prize that motivates a candidate’s behaviour is building local influence and accessing local rent opportunities, the candidate can avoid the large national parties and still have a lucrative political career. As such, in electoral districts with high rent opportunities, viable candidates are more likely to join minor parties. 38 Hicken’s formal presentation is: EUlarge = p(EPL) + (1-p)(EP~L) – C where EUlarge is the payoff to a candidate for coordinating their efforts with a large party across districts, P is the probability the coordination produces the largest party, and 1-p is the probability that coordination produces a party that is not the largest. EPL is the payoff for being the largest party, and EP~L is the payoff for being a party that is not the largest. C is the costs incurred for cross-district coordination. 35 Candidate-centered campaigns The issue of candidate campaigns follows closely from the issue of affiliation. Campaigns involve efforts by the candidate to persuade the electorate to offer their votes, whether through programmatic or clientelistic promises. There are many different ways to campaign; however we can broadly separate campaigns based on their emphasis of either party-centered or candidate-centered appeals. Why do candidates emphasize personal over partisan messages? Consider the local campaign in an area with high rent opportunities. If voters are primarily concerned with ensuring access to individualized benefits from the local government, national policy concerns will be minimal. The major party labels communicate very little about a local candidate’s ability to deliver rewards once in office. Given that competition across parties on issues is less likely to be relevant to the average voter, candidates are driven to compete by either pledging their ability to deliver goods once elected and/or pointing out their past success at this task if they are an incumbent. Campaigning takes the form of individualized or club goods delivery. A new set of dishes or the repair of a local temple demonstrates a candidate’s willingness to deliver particularistic goods to her targeted supporters once in office. Voters and candidates struggle with time-inconsistency issues: goods delivered by the candidate before election do not guarantee votes, and the election of a purportedly friendly patron does not guarantee a voter goods delivery once the patron has been elected. The more relevant point, however, is that these personal, patronage-centered appeals are likely to take place separately from national-level party competition over policy and governmental control. In a campaign where the promise of particularistic goods delivery is essential, a candidate running under a minor party label can compete with a candidate from a major party if voters anticipate that even minor party politicians will have some influence over the distribution of valuable local goods after an election. Rent opportunities and party system size All three candidate-level effects of rent opportunities – increased entry, affiliation with minor parties, candidate-centered campaigns – combine to fragment the party system. The basic mechanism works through minor party support. Minor parties in high rent areas are more likely to have slates with a respectable number of viable candidates. These 36 locally viable minor party candidates appeal to the electorate by promising to deliver particularistic goods once elected. Consequently, the vote share for minor parties’ increases at the expense of the major parties. Yet the major national parties are not entirely routed by their minor competitors. They are able to tap into that portion of the electorate that retains a national orientation. Also, they adapt by recruiting candidates that can compete with clientelistic appeals. In high rent districts, the major parties lose vote- share but they are not completely displaced by minor parties. Rather, the success of minor parties expands the number of parties in the local party system. Scope conditions and contributions The dissertation links rent opportunities to party system outcomes. The immediate applicability of the theory is bounded by assumptions being made about the formal and informal institutional environment. In this section I emphasize the contributions made by my research while noting scope conditions that bound my theoretical model. Rent sharing expectations The theoretical arguments linking rent opportunities and party system fragmentation depend on the condition of local rent sharing. By local rent sharing I refer to expectations that candidates and voters hold regarding the control of rents in a sub-national system. Local rent sharing is high when voters and elites expect parties to share in the distribution of rents rather than exclude one or more major actors. Where there are high expectations for local rent sharing, voters and elites expect to see universal or near-universal governance coalitions. Where there are low expectations for local rent-sharing, voters and elites anticipate rents will be controlled by either a large party or a minimal winning coalition of parties. In political systems with high-levels of rent sharing, there are no losers; in systems with low-level of rent sharing, actors are excluded from accessing rents. This raises the question: why would there ever be expectations of rent sharing? Studies of legislative coalition formation provide potential answers, typically hinging on the information constraints facing legislators and party leaders. From the point of view of the legislator, coalition behaviour is strongly conditioned by whether politics is 37 dominated by the provision of either distributive goods or public goods.39 If a politician cares about re-election, and access to distributive goods increases the chance of re- election, there are reasons to prefer a universalism norm. The legislator may receive more benefit from inclusion in a minimum winning coalition; however, they would rather guarantee access to distributive goods in a universal coalition than risk being cut out of a minimal winning coalition. From the point of view of party elites during coalition formation, leaders may prefer to have more parties than necessary if they cannot rely on the loyalty of the coalition members. Extra coalition members ensure a small numbers of defections will not forestall government initiatives (Volden and Carrubba 2004). While this game theoretic logic has not been applied to the Indonesian case, the country’s coalition politics neatly fit existing empirical predictions. Universal, or near universal, coalitions are the norm in Indonesia. Indonesian politics are primarily about distribution (often in the form of corruption) and no politician wants to be cut off from the spoils. When politics switches to questions of public goods provision, legislative committees are notoriously independent and coalition leaders cannot reliably depend on their legislators to toe the party line. Legislator-level desire to maintain access to rents as well as leader-level concerns about government stability creates the incentives for over- sized coalitions. Yet these are not the only explanations for Indonesia’s universalism norm. In his explanation of over-sized coalitions in Indonesia, Slater (2004) emphasizes the importance of retaining access to state resources. However, he goes further and suggests the cartelistic behaviours can be partially explained because no contender expects an electoral breakthrough in which they would rule the legislature. Additional arguments can be made emphasizing the impact Indonesia’s formal and informal institutions have on coalition behaviour. 40 The important point is that most parties that win seats are able to access rents in Indonesia, even small ones. 39 Distributive goods have concentrated benefits, diffuse costs, and can be disaggregated unit by unit. Public goods are provided to all citizens equally (Weingast 1979). 40 Slater also notes (and dismisses) the cultural arguments used by Indonesian leaders themselves to justify their over-sized coalitions (2004, 66). He is correct that national cultural traditions do not necessitate elite collusion. But popular cultural concepts like “mufakat” (consensus) and “gotong-royong” (mutual help) make it easier for politicians to publically justify their decision to cooperate with erstwhile rivals. Even if informal codes of conduct do not directly cause Indonesia’s over sized coalitions, it is plausible that they help sustain the practice in light of public scrutiny. 38 Table 2 - Party system size: theoretical predictions Scope of Local Rent Opportunities High Rents Low Rents Local Rent Sharing Expected? Yes No A. Many Parties B. Fewer than Cell A C. Fewer than Cell A D. More than Cell C When candidates expect local rent sharing in high rent political systems I hypothesize they alter their behaviour in ways consistent with my model. However, where candidates do not expect local rents sharing, my theoretical predictions should not hold. Table 2 presents a 2X2 with empirical predictions regarding number of parties by extent of rent opportunities and expectations of rent sharing. High rent opportunities produce party systems that have either a relatively high or relatively low number of parties, depending on expectations of rent sharing. The combination of no rent sharing expectations and high rent opportunities should create the conditions for the success of a machine party. The scope condition limits the applicability of the theory to only political systems with expectations of rent sharing, but it also places the work into comparative perspective. It is likely that rent opportunities facilitated the growth and maintenance of party machines in some political systems. My work suggests conditions under which rewards are distributed by either party machines or personalistic networks. Ethnic parties and ethnic party bans A second scope condition relates to the non-existence of ethnic parties. My argument suggests that ethnic diversity can shape the party system through a mechanism more subtle than the ‘ethnification’ of parties. Indonesia is a case where parties are effectively blocked from forming along ethno-linguistic lines; nonetheless, ethnic structure can still affect long-term patterns of sub-national spending and these patterns affect the incentives offered to both voters and elites. By altering the political-economic context, ethnic diversity manages to shape party system outcomes without necessarily producing an ethnic party system. Where political actors are not prohibited from launching and voting for ethnic parties, it is unlikely that my theoretical story will hold. The existence of clear ethnic 39 parties facilitates processes of communal polarization which alter the strategies of elites and voters in ways that disrupt my causal story. While my proposed mechanism only works in countries where ethnic party formation has been prohibited, awareness of this more subtle mechanism should still inform future cross-national studies of ethnic diversity and party systems. The prohibition of ethnic parties has become a common tool of institutional design.41 The considerable literature that has built up around the exploration of ethnic diversity and party system size has not considered the issue of ethnic party bans; indeed, only recently has the proposed mechanism between the ethnic diversity and the multiplication of ethnically based parties come under scrutiny.42 Though the application of the theory is bounded by institutional conditions, this close study of politics in one country with an ethnic party ban has implications for a broader universe of cases. Decentralization and sub-national power A third theoretical scope condition relates to assumptions made about the distribution of power across different levels of government. Previous theorists have highlighted the way vertical and horizontal centralization of authority can shape national party system outcomes. They have not applied the approach to understanding variations across electoral districts. This dissertation breaks new ground by extending the insights of the aggregative approach down to the district-level. Indonesia offers an interesting case where the competition for both provincial and municipal power is a significant factor influencing national-level voting decisions. The specific theoretical framework outlined above depends upon two institutional conditions. First, sub-national units have effective power to distribute goods that are valued by the electorate. Second, elites are required to join nationally certified parties in order to compete for sub-national office. While these two conditions do not apply to all countries, what is broadly applicable is the concept that the units of governance contained within an electoral district can potentially shape both patterns of party organization and party system outcomes. Where municipalities have a 41 Though instances of ethnic party prohibition in sub-Saharan Africa have received the most scholarly attention, they have also been enacted in South Asia, the Middle-East, and several post-communist states (Basedau, Bogaards, and Hartmann 2007). 42 Chandra, for instance, has begun a major project to measure the ‘ethnification’ of party systems (Chandra and Wilkinson 2008). 40 significant impact on voters’ lives, the struggle for municipal power can potentially percolate up, affecting how voters and elites align themselves in national-level competitions. 41 Chapter 3 - Indonesia Background This dissertation is concerned with the relationships between ethnicity, rent opportunities, and party systems. While Indonesia’s institutional framework was set up to prevent the formation of regional and ethnic parties, we still see the residual effect of ethnic diversity on the party system. Specifically, ethnic diversity shapes the local rent-seeking opportunities open to politicians, which thereby alters patterns of political participation by elites and voters. The empirical chapters link party system size to rent opportunities through an examination of party organization and electoral campaigns. The theoretical story takes place in the complex institutional and social context of the Indonesian case. Some case-level knowledge is important before discussing the issue of party system evolution. Key questions need to be addressed: Why have Indonesian institutional designers tried to limit the partisan expression of sub-national identities? Beyond rent opportunities, which institutions and cleavages shape the Indonesian party system? What are the key indicators of rent opportunities? Why do rent opportunities correlate with ethnic diversity? The background proceeds in four sections. The first section examines the political challenge posed by ethnic diversity in Indonesia. I trace the unease with ethnic parties to the development of nationalist norms. The second section describes the evolution of the country’s political institutions that directly shape the Indonesian party system. Particular attention is paid to the legal tools used to suppress regional and ethnic party formation. In the third section I describe the major players, the cleavage structure, and the coalition patterns that typify the contemporary Indonesian party system. The fourth section examines rent opportunities across Indonesia. I describe how decentralization increased the material and political importance of sub-national power. I then link variation in local rent opportunities to ethnic diversity. The origins of nationalist norms Efforts to prevent regional and ethnic party formation in Indonesia have been motivated by a concern for national unity. One only has to look at a map to gain some appreciation for the challenge of managing diversity and maintaining territorial integrity. It seems only natural for an archipelagic country to remain on guard against possible separatist islands. 42 Still, part of the country’s allergy to ethnic parties can be traced back to its particular brand of nationalism. To understand the norms that shape contemporary political institutions it is useful to revisit the ethnic and colonial context that confronted Indonesia’s early nationalists. Ethnicity in Indonesia Indonesia is a country of more than 10,000 islands that contain people speaking over 700 different languages. Like many post-colonial countries, Indonesia is a product of the colonizer. The country’s present borders follow the limits of the old Netherlands Indies, a colony that was shaped by competition with imperial rivals more than it was the contours of human or physical geography. Peculiarities abound. Indonesia claims only the southern portion of Borneo and the western portions of Timor and New Guinea. Culturally Malay areas of eastern Sumatra are separated by international border from co- ethnics on the nearby Malay Peninsula. Although the country is predominantly Muslim, there are islands dominated by Hindus, Protestants, and Catholics. As one Western observer of the country suggests, Indonesia itself is an ‘unlikely nation.’43 Ethnic groups in Indonesia are officially referred to as suku bangsa (sub-nation), often shortened to simply suku. The official term suku bangsa implies a sub-national group cannot ever reach the level of bangsa (nation) and thus contains an assumption about the natural order of political organization. Rebellious groups such as the Acehnese have often referred to themselves as a bangsa, a rhetorical move to indicate the group’s self-confidence and claim to political autonomy (Aspinall 2006, 174). The Acehnese claim to nationhood does overlap with the relatively homogenous Acehnese ethnic group. Complicating the situation further, the regionalist movement in Papua refers to itself as a bangsa that contains multiple suku bangsa. Despite the contested political nature of the term, I will refer to all sub-national descent based identities interchangeably as either ‘ethnic groups’ or suku. The limited definition excludes the regionalist claims to nationhood such as Bangsa Papua. 43 Technically, title of “Indonesia: The Unlikely Nation?” leaves the question open (C. Brown 2003). 43 Unlike the categorization of religion, the Indonesian state does not follow rigid bureaucratic guidelines for the categorization of citizen’s ethnicity.44 It was not until 2000 that the Indonesian authorities finally published census data that included ethnic composition. Dutch efforts in 1930 sorted people by ‘social criteria,’ namely language and custom. Census-workers identified 137 groups in that year, though the number was substantially circumscribed due to census-takers relying on their assessment rather than self-reporting (van Klinken 2003, 101). In 2000, census-workers were more sensitive to self-identification and individuals were coded as belonging to one of 1,072 different groups. Most groups are small. Papua alone contains hundreds of different suku, but only two indigenous groups – Biak Numfor and Dani – contain over 100,000 people. At 42% of the population, the Javanese constitute Indonesia’s largest ethnic group. The only other group with over 10% of the population is the Sundanese (15%). The demographic weight of the plurality group frequently sparks accusations of Javanese chauvinism. Given that the Javanese are concentrated on one island – Java – complaints are often voiced in geographic terms. Indonesian policies are targeted by those from the Outer Islands for ignoring the needs of the ‘regions.’ Presidential tickets commonly attempt to balance a Javanese candidate with an Outer Islander. Also, the presence of Javanese migrants can motivate a ‘native sons’ (putra daerah) political backlash.45 Despite the existence of anti- Java rhetoric, though, there is nothing approaching a common Outer Island identity. Observers occasionally fear the break up and ‘balkanization’ of Indonesia (Hadar 2000; Wanandi 2002). The borders have, however, remained relatively stable; the only successful secession took place in East Timor, a province that was never part of the original Dutch colony. While territorial integrity has sometimes been enforced at the barrel of a gun, coercion alone cannot explain Indonesia’s endurance. The stability of the 44 Indonesia has long had five official religions (Islam, Catholicism, Protestantism, Hinduism, Buddhism). All Indonesians must proscribe to one of the options, which appear on a citizen’s identification card. The practice leaves little opportunity for self-expression, papering over substantial internal diversity within the Muslim community as well as forcing followers of indigenous faiths to identify with a religion of which they may have little knowledge. 45 It is not just Javanese migrants that motivate responses from local groups, though the raw number of Javanese migrants makes them the most likely source of backlash. 44 country’s borders serve as one indicator of the resilience of Indonesia’s brand of non- ethnic civic nationalism. The rise of Indonesian nationalism While Indonesian textbooks suggest the existence of pre-colonial unity, most observers trace the conditions of Indonesian nationalism back to Dutch colonial policy. According to Henley, “Until the consolidation of the Dutch colonial state, Bataks and Balinese, if they had heard of each other at all, had no more notion of sharing an Indonesian - or any other - identity than did Tagalogs or Merinas” (Henley 1995, 289). Anderson highlights a few crucial colonial policies that forged the Indonesian identity. Classroom maps of the Netherlands Indies imprinted the idea of the nation in the imagination of students (Anderson 1991, 176-178). The official promotion of Malay provided a common vernacular (Anderson 1991, 131-132). As well, the development of an integrated colonial bureaucracy strengthened a common sense of identity among the elite (Anderson 1991, 121-122). These factors provided the foundation on which Indonesian nationalism would later develop As Henley points out, though, Indonesian nationalism was by no means inevitable. Several policies highlighted by Anderson were also present in French Indochina, which later fragmented into three countries. For Henley, Indonesian nationalism is explained by the non-emergence of Javanese exclusivist nationalism. Java’s weak historical memory of national unity, combined with the gradual pace of Dutch colonialism, stunted the development of an exclusivist Javanese nationalism. Had the Javanese experienced both centuries of pre-colonial rivalries with foreign powers and the swift arrival of a colonial power, the group would have possessed a stronger sense of identity that could have led to either the political fragmentation seen in French Indochina or the violent ethnic chauvinism seen in Burma. Instead, the non-emergence of Javanese nationalism opened space for an inclusive, civic nationalism. From the outset of the nationalist movement, ethnic divisions were an obstacle to national unity. Sukarno, according to Legge, ‘deplored’ the divisions generated by the 45 patchwork of religious and ethnic associations (Legge 2003, 91).46 After the founding of Indonesian Nationalist Party (Partai Nasionalis Indonesia, PNI), Sukarno and his allies assembled representatives of the ethnic associations in order to win support for the movement. The meeting resulted in the famous ‘Youth Pledge’ of ‘One Fatherland, One Nation, One Language’ (satu nusa, satu bangsa, satu bahasa), a clear repudiation of political organization around sub-national identities. Nationalist suspicion of sub-national identities grew during the Revolutionary period (1945-1949) when, in a bid to win local support, the Dutch created a considerable number of ‘states’ in the areas they controlled. In many cases these puppet states followed ethnic lines.47 The nationalist Republican government accused the Dutch of ‘divide and rule’ tactics and nationalists were expected to reject the creation of Dutch ethnically based units and agitate for the unitary Indonesian Republic. The eventual victory of Republican forces bolstered nationalist identity. In a nod to internal pluralism, Indonesia proclaimed its motto Unity in Diversity (Bhinneka Tunggal Ika). Nonetheless, the realities of governing an under-developed, post-colonial state brought the regional and ethnic tensions to the fore. The Republic soon faced a series of regional revolts. The response from the centre was to further stigmatize sub- national loyalties. Ethnicity became a taboo subject due to its ‘explosive potential’ (van Klinken and Nordholt 2007, 22). Sukuisme, defined broadly as forwarding demands for the benefits of one’s own group, became a common pejorative. During the Suharto years, the press was regularly lectured on avoiding any discussion of ethnicity, religion, race, and class (Suku, Agama, Ras Antar Golongan, SARA). In dealing with ‘ethnic’ questions, Indonesia has lived a double life: the concept of diversity could be celebrated in an official capacity but strong norms built up against the honest discussion of ethnic tension. Vigilance against signs of SARA persists, especially among the urban elite. Despite the recent liberalization of the political sphere, these norms have been embedded in the country’s electoral institutions and party programs. 46 Ironically, Sukarno started his political career in “young Java”, an organization of the type he would later seek to sideline. 47 ‘Pasundan’ was formed in Sundanese areas, ‘Madura’ in areas with Madurese, ‘Great Dayak’ in Dayak areas. For more on Dutch policy during the nationalist revolution, see: (Kahin 1952, 351-390). 46 Indonesian political institutions The evolution of Indonesia’s party system has been affected by the institutional context. The introduction of direct presidential elections encouraged the entrance of new partisan actors. Electoral laws have ensured only certain types of parties make it to the ballot. In this section I describe the evolution of Indonesia’s electoral laws and its system of executive-legislative relations. First, I trace the origins of Indonesia’s current presidential system back to its founding constitutional document. In discussing the evolution of executive-legislative powers, I briefly touch on the rise and fall of Indonesia’s past regimes. Second, I provide a detailed description of Indonesia’s electoral system. I give close attention to three underlying trends in electoral reform: 1) increasingly stringent party system nationalization laws; 2) efforts to curb the number of parties; 3) reforms to improve constituent-legislator connections. Executive-legislative structure Constitutional continuity and regime change: From Independence to the New Order Indonesia is one of the few countries to shift from presidentialism to parliamentarism and back again. The country’s first constitution, Undang-Undang Dasar 1945 (UUD ’45, Basic Laws of 1945), was written under the aegis of a sympathetic Japanese military. UUD ’45 granted wide powers to the executive and created a weak legislative branch. Politically, the office of the executive was largely crafted to accommodate Sukarno, the revolutionary leader and symbol of the new Indonesian nation. The president was to be selected every five years by a supra-legislative body known as the People’s Consultative Assembly (Majelis Permusyawaratan Raykat, MPR). The MPR was to be composed of members of the national legislature, known as the People’s Representative Council (Dewan Perwakilian Rakyat, DPR), as well as appointed representatives from regional and societal groups. Despite being selected by the MPR, the president was not accountable to any legislative body. The final document remained ambiguous, setting up an executive structure that has been called “presidential with parliamentary characteristics” (McIntyre 2005, 6).48 48 Beyond assigning powers to the different branches of government, the constitutional preamble in UUD ‘45 established Pancasila as the official national ideology. Pancasila (Sanskirt for ‘five principles’) 47 The return of Dutch colonial troops and the need to gain international recognition shifted the balance of power within Indonesian political society. Political momentum fell to the loosely connected of youth groups (pemuda) who had resisted the Japanese and were bent on confronting the Dutch. These activists successfully pressured the Republic’s leadership to adopt a de-facto parliamentary system in which the cabinet would be accountable to a legislative body. The agreement put in place institutions that would last more than a decade.49 Cabinets in this period rose and fell with the support of the legislative body. Even though elections were not held until 1955, broad partisan representation within the legislature ensured that the parliamentary game was viewed as the legitimate focal point of political life. In 1957, a series of regional rebellions created a crisis that brought down the governing coalition. Sukarno declared a ‘State of War and Siege.’ Following unsuccessful attempts to form a government, Sukarno took the role of formateur on himself and constructed a cabinet that circumvented the power of the legislative parties. The increasingly assertive president then proposed returning to UUD ’45, a move that would substantially increase his formal powers. When the parties would not accede to his demand, he instituted the constitutional change by decree. The experiment with parliamentary governance was brought to an end and Indonesia entered an era of ‘Guided Democracy’ (Demokrasi Terpimpin). Sukarno sought to balance competing forces against each other. Rivalries that positioned the president at the centre of the political spectrum were nurtured. The most precarious strategy involved balancing the increasingly assertive Armed Forces with the growing Indonesian Communist Party (Partai Komunis Indonesia, PKI). In 1965, a coup attempt by armed service personnel with ties to the PKI triggered a violent response by encompasses a commitment to a series of abstract social and political values, including monotheism, national unity, humanitarianism, deliberation and consensus, and social justice. The emphasis on Pancasila, and the last minute omission of any reference to religious practice, caused controversy among Islamists, who were disappointed the constitution did not oblige Muslims to follow Islamic law. In subsequent years, Islamist politicians would push for constitutional reforms that would enshrine Muslim religious practices. The constitutional conflict between supporters of the relatively secular Pancasila ideology and supporters of Islamic law contributed to the eventual collapse of the democratic regime in the late 1950s. 49 The departure of the Dutch and the unification of Indonesia led to provisional constitutions in 1949 and 1950, though the parliamentary character of the state remained intact. 48 anti-communist forces.50 Sukarno’s ties to the coup leaders implicated him the plot. The military, led by General Suharto, slowly marginalized Sukarno and eventually compelled the president to transfer authority and step down from his post. Notably, the transfer of power took place within the context of UUD ’45, thereby providing a veneer of legality to the slow-moving military takeover. In 1968 Suharto was officially made President, formalizing the authority he had managed to artfully appropriate. During Suharto’s reign, cabinet positions and top military commissions were the focal points of political competition. Competition for positions was analysed through the lens of factional conflict: secular nationalist “red and whites” jostled with conservative Muslim “greens,” Outer Islanders struggled against perceived Javanese supremacy, civilians attempted to expand their power at the expense of the military, and technocrats fought rear-guard battles against Suharto’s cronies. The factions were never organized per se, and high ranking officials played different roles at different times. Like his predecessor, Suharto tacitly encouraged these disputes as they allowed him to divide his potential challengers and tip the balance between different policy views as he saw fit. Despite the occasional appearance of internal division, Suharto went beyond Sukarno in the consolidation of policy-making authority in the office of the executive. The legislature was reduced to a largely symbolic role. Legislator’s fulfilled their assigned role in the pageant and most faithfully followed the “5-Ds”: Datang, duduk, dengar, diam, duit (“Turn up, sit down, listen, shut up, get paid"). By the late 1990s, Suharto and the New Order regime he had created were showing signs of ageing. The long taboo subject of succession was broached with increased frequency. Positions of power were routinely dolled out to family and trusted cronies. Even the pliant opposition parties showed signs of defiance. When the Asian financial crisis hit Indonesia it exposed the weaknesses of Suharto’s regime. For months the President vacillated in his response. His 1998 decision to drastically reduce subsidies on household goods provided the trigger for mass protests aimed at the regime and Suharto himself. Faced with strong outside pressure, key allies abandoned Suharto. In May 1998 Suharto resigned and Jusuf Habibie, the recently installed Vice-President, assumed the position of Presidency and quickly launched a series of democratic reforms. 50 For a concise review of the post-coup conflict, see: (Crouch 1988, 135-157). 49 Constitutional Reform in the Reformasi era51 In June of 1999 the country held its first free election since 1955 and the process of constitutional reform began shortly thereafter. The early rounds of executive-legislative reforms responded to the perceived failings of the Suharto era.52 In the first amendment round the presidency was stripped of its right to make law without consent from the DPR, executive office-holding was limited to two-terms, and the president’s pardon and diplomatic rights were curbed. In the second round, the military lost its reserved seats in the DPR, depriving an organization responsible to the executive branch of representation in the legislature. Further reforms eliminated the president’s ‘pocket veto’ and strengthened the DPR’s supervisory powers. Taken together, the first and second amendments made the once weak DPR a centre of considerable power. UUD 45’s vague separation of powers set in place the conditions for a constitutional crisis. By 2000, the recently installed President Abdurrahman Wahid had managed to alienate the legislative coalition that had brought him to power. The president and his legislative rivals launched corruption investigations against each other. Pitched battles over cabinet positions and the appointment of security officials continued for months. After repeated censures and a presidential threat to dissolve the legislature, the MPR finally impeached Wahid in July 2001. Wahid’s impeachment provided the backdrop for further constitutional amendments. Some of the reforms were immediate responses to the crisis. The impeachment process was clarified, providing an expanded role for the new Constitutional Court. The legislature’s separate mandate was also clarified by adding an Article stating that the President could not dissolve the DPR (as Wahid has attempted), and the issue of Vice-Presidential selection in case of vacancy was cleared by explicitly granting this power to the MPR. The process of presidential selection was also reformed. After persistent resistance by the Indonesian Democratic Party of Struggle (Partai Demokrasi Indonesia – Perjuangan, PDIP), the legislature’s largest party, the MPR introduced a direct 51 The use of ‘Reformasi’ here refers to the post-Suharto democratic era. The term can also refer to the social and political movement that spearheaded the transition to democracy. 52 The following account is based on Denny Indrayana’s Indonesian Constitutional Reform 1999-2002: An Evaluation of Constitution-Making in Transition. (Indrayana 2008) 50 presidential contest. Only parties achieving a threshold of support in the legislative election would be permitted to nominate a joint Presidential-Vice Presidential ticket.53 The presidential elections, then, would have to occur after the legislative competition. Aspiring presidential candidates would require legislative allies to even run, creating an incentive for potential candidates to start their own parties. The conditions placed on presidential victory were demanding: a first round victory required a ticket to secure over 50% of the total vote, with a minimum of 20% of the vote in over 50% of the country’s provinces. A run-off ballot would determine the winner in the event that no team reached the demanding threshold. Another outcome of the post-Wahid constitutional reform was the creation of a new legislative body named the Regional Representative’s Council (Dewan Perwakilan Dareah, DPD). The DPD was created with the purpose of strengthening the voice of the ‘regions’ in Jakarta. The new chamber was granted a mix of oversight and advisory powers, though it lacked budgetary authority and was only permitted to submit bills to the DPR pertaining to a limited number of subjects.54 In addition, members of the DPD would sit as representatives in the MPR. Unlike the DPR, which was restricted to representatives with partisan backing, the DPD was to represent ‘individuals.’55 The stipulation effectively barred organized partisan campaigns for DPD seats. As Indonesia approached the 2004 elections its executive-legislative structure looked substantially different than it had a mere five years before. UUD ’45, a document that had structured two authoritarian eras, was brought into line with the country’s emerging democratic norms and shorn of its parliamentary characteristics. The removal of non-elected elements from the legislature, the introduction of direct presidential elections, and the creation of a directly president substantially increased the direct control of the voters over the composition of the legislature and executive. 53 Article 6A (2). The threshold for participation was not specified in the amendment. 54 The powers of the DPD are described in Article 22D. The DPD can only submit bills to the DPR that deal directly with issues of regional governance, such as the fiscal balance between the centre and the regions and the formation of new regional governments. On other keys issues like education, religion, and state budgets, the DPD can only advise the DPR and supervise implementation of law. 55 Article 22E (4). 51 Electoral systems and party laws All of Indonesia’s legislative elections have taken place under a system of proportional representation. Feith notes that the initial adoption of PR in the 1950s occurred with no serious opposition (1957, 3). Besides the fact that PR was the system most familiar to young nationalists that had studied in the Netherlands, the system no doubt appealed to the multitude of partisan actors unsure of their electoral strength. Experience with a fragmented legislature did prompt some non-partisan actors to re-evaluate the PR system. Plans to move to a single-member district plurality system were formulated prior to the first New Order election while a mixed SMD-P/PR system was proposed by an influential committee in 1998 (Budiardo 2001, 127-131). Nonetheless, the proposals to replace PR never found a partisan champion. The desired ends of electoral system design go beyond achieving a broadly representative legislature and a close approximation of seats to vote share. There are at least three additional goals that have been relevant to electoral system designers: 1) party system nationalization; 2) a moderate number of parties; 3) close legislator-constituent relations. These goals are reflected in electoral and party law and have had an impact on party system evolution. Party system nationalization The Indonesian electoral system design has been shaped by concern for the integrity of the country. During the country’s first stint with parliamentary government there was a rash of regional rebellions.56 Regional concentrations of partisan support were thought to have fuelled the disputes. In particular, the disproportionate strength of Masyumi support in the Outer Islands made the party one of the few tenuous links between Jakarta and the regions.57 When Masyumi was marginalized in the capital, regional feelings were strengthened.58 The situation underlined the danger of a regionally based party system. 56 In most cases the conflicts were not clearly separatist. Rather, they were regionally concentrated revolts against the central government. Both the Darul Islam movement and the Revolutionary Government of the Republic of Indonesia (Pemerintah Revolusioner Republik Indonesia, PRRI) sought to overthrow or replace the government rather than form breakaway states. In contrast, the uprising in Ambon was explicitly separatist in motivation. 57 Masyumi was an acronym for Majelis Syuro Muslimin Indonesia (Council of Indonesian Muslim Associations). The party forwarded a modernist Islamic platform. 58 Several high ranking members of Masyumi later joined the provisional government set up by the PRRI. 52 Another wave of regional and ethnic conflict swept the country with the fall of Suharto. Separatist movements gathered strength in Aceh and East Timor. Ethnic and religious groups fought in Kalimantan, Central Sulawesi, and the Malukus (Bertrand 2004). Ethnic Chinese found themselves targeted in numerous locations. The violence clearly demonstrated that decades of suppressing conflict with an authoritarian hand had not created a harmonious society. Concern for national integrity manifested itself in regional requirements on partisan organization. Regional requirements were first put in place by Sukarno. Early in the Guided Democracy era, the president forwarded a law that parties had to organize branches in a quarter of all provinces and a quarter of all municipalities, effectively prohibiting regional and ethnic parties (Legge 1961, 224-5). During the New Order party registration was tightly controlled, with authorities allowing for the existence of only three national parties that were required to compete across the country. In the reformasi era a new legal framework for political parties was established by the passing of Law No. 2 of 1999 on Political Parties. Few constraints were placed upon party formation, though nationalist norms were apparent. For instance, Article 3 did include a provision prohibiting parties from endangering the integrity and unity of the nation, while Article 2 (c) required parties to allow all Indonesian citizens the right to become members. While the language indicated an attempt to discourage parties forming along regional and ethnic lines, the enforcement mechanisms were left vague. Law No. 31 of 2002 on Political Parties tightened the rules for party registration. Whereas the 1999 law allowed parties to form with 50 signatories, in 2002 a party would need to demonstrate the ability to organize in 50% of municipalities in 50% of all provinces simply to register.59 Political Party law in 2008 raised the regional requirement to 60% of all provinces.60 Regional requirements were even stricter in general elections laws. In early drafts, all parties participating in the election were required to demonstrate an organizational presence in half of all municipalities in half of all provinces.61 The requirement 59 Law No. 31/2002, Article 2 (3) 60 Law No. 2 / 2008, Article 3(2) 61 Law No. 3/1999, Article 39 (1) 53 effectively prohibited regional parties from competing in even sub-national election. In the face of criticism the government added a transitional provision allowing parties organized in only 1/3 of all provinces to compete in 1999.62 In 2004 the requirement was raised to 2/3 of all provinces.63 Regional requirements for either party registration of electoral participation have tightened with each election. It is striking how little opposition existed to such sweeping regional requirements. Reflecting on the issue, Donald Horowitz notes, “The rationale for such a widely distributed support can only be described in terms of a severe allergy to small parties and to regional or ethnic parties” (Horowitz 2001, 140). Even in 1999, the opposition had more to do with fairness for new parties that were only recently granted the right to organize than a concern for regionalist movements (Nasution 2001, 134). The acceptance of such rules indicates the existence of strong nationalist norms at the elite level. Containing Party Numbers Despite Indonesia’s commitment to PR, a system that enables the success of minor parties, the country’s institutional designers have repeatedly tried to curb the number of political competitors. Efforts to contain the number of parties stretch back to Sukarno, who was particularly strident in his criticism the party system. In his famous ‘Bury the Parties’ speech of 1956, Sukarno stated: There is a disease that is sometimes even worse than ethnic and regional feelings! What is this disease, you ask? It is the disease of parties, brothers and sisters! Yes, I will be frank: the disease of parties (Feith and Castles 1970, 81). Parties, according to Sukarno, were self-serving and out of sync with Indonesia’s consensual culture. Their existence created societal discord. Part of the problem, in Sukarno’s view, lay in the sheer number of partisan actors. He stated: As you know, in one of my Independence Day Speeches made more than a year ago, I said I hoped the general elections would be able to restore our party system to health. Remember, at that time I said I hoped the elections would be able to reduce the number of our parties, which at that time stood at thirty, so that there would be just a few parties. This is what I was hoping!...But, look what happened! 62 Law No. 3/1999, Article 82. 63 Article 7 (1) 54 After the elections there were even more parties than before, even more. (Feith and Castles 1970, 82-3) Like many political observers in his day, Sukarno linked the problems of governance directly to the perceived excess of parties. He also took action, banning parties and introducing strict party registration requirements. Sukarno’s critique of the party system was picked up by Suharto during the New Order. For Suharto, partisan conflict created unnecessary confusion. There was only one valid political goal – development – and there could only be minor differences in how to achieve the goal. The President used a transportation analogy to explain his logic: [W]ith the one and only road there, why must we have so many cars, as many as nine? Why must we have wild speeding collisions?...It is not necessary to have so many vehicles. But it is not necessary to have only one. Two or three is fine. (Elson 2001, 189) In order to simplify the system, the regime forced the merger of all existing parties into three camps (nationalist, spiritual, functional) representing different aspects of Suharto’s vision of the developmentalist agenda (Elson 2001, 190) The persistent concern with an excessive number of parties carried into the reformasi era. The proportional system, combined with a loose party registration process, opened the door for party system fragmentation. The decision to open up the electoral competition to a multitude of actors was deliberate. After 35 years of controlled elections there was a strong demand for substantial liberalization. Yet even during the transition period the institutional designers were putting in place laws that would consolidate the party system of the future. Specifically, Law No. 2/1999 ensured that parties not reaching a threshold of 2% of national seats in 1999 were barred from participation in the 2004 election.64 The party registration threshold created pressure on the minor parties to combine their efforts or drop out altogether. In 2003, the party registration threshold was raised from 2% of all seats to 3%.65 Institutional designers changed tactics in the run up to 2008. A new legislative threshold was introduced barring parties receiving less than 2.5% of the national vote 64 See Law No.3/1999, Article 39 (3). Parties were also permitted to run if they received 3% of all provincial seats or 4% of all municipal seats in 50% of all provinces or municipalities. 65 Law No. 12/2003, Article 9 (1a) 55 from seating legislators.66 As a concession to the minor parties, the restrictions on party participation were eased. Whereas the law of 2003 stipulated that only parties that received 3% of the seats in the national legislature had the right to participate in the next election, the 2008 law opened competition to all those receiving at least 1 seat in 2004.67 Nonetheless, the loophole was specifically written to only apply to one election and the participation restrictions from 2003 were otherwise left in place. Legislative and party registration thresholds were not the only pressures placed on minor parties. Electoral district reform reduced the probability minor parties would win legislative seats. In 1999, provinces were allocated a certain number of seats and electoral districts followed provincial boundaries. Consequently, large provinces had high district magnitudes. The allocation of seats in these districts closely matched the distribution of votes, thereby benefitting minor parties. In 2003, the electoral law prevented an electoral district from exceeding 12 seats.68 Though the law was reportedly written to close the gap between legislators and constituents, the large parties were no doubt aware of the advantage provided by small district size.69 Prior to the 2009 election the maximum number of seats in a district was lowered from 12 to 10.70 In sum: Since the parliamentary era of the 1950s there has been a deep-seated aversion for party system fragmentation. The preference for a moderate number of parties sits uneasily with the national commitment to proportional representation. Nonetheless, there have been numerous reforms to party and elections laws designed to pare back the number of parties. Party registration rules, a legislative threshold, and shrinking districts are all consistent with a desire to prevent an excess of parties. 66 Law No. 10/2008, Article 202 (1) 67 Law No. 10/2008, Article 316 (d) 68 Law No. 12/2003, Article 46 (2) 69 The issue of large party advantage was addressed in both the popular press and KPU news releases. See Kompas August 16 2003, “Small-Medium Parties will be ended by the veiled threshold” [Parpol Kecil- Menengah Akan Habis oleh “Threshold“ Terselubung]; January 8 2004, “The 2004 Election: Death Knell for Small Parties” [Pemilu 2004: Lonceng Kematian Partai Kecil]; KPU, “The Electoral System According to Law 12. of 2003” [Sistem Pemilu Menurut No. 12. Tahun 2003]. 70 Law No. 10/2008, Article 22 (2) 56 Improving legislator-constituent ties A third stream of institutional reforms involves efforts to increase legislator accountability to his/her constituents. In 1955, Indonesia’s first electoral system offered literate voters a chance to cast a preference vote by writing in the name of a candidate from a party’s candidate list. This minimal amount of control was lost during Suharto era elections. New Order officials closely screened candidates to ensure criticism of the regime would be restrained. Candidates for office achieved their spots due to party loyalty and network connections. Many candidates ran in districts in which they had no meaningful connections. Since the transition, reformers have searched for ways to close the perceived legislator-constituent gap. The first method, used in the 1999 election, injected a measure of plurality competition into the over-arching framework of PR. Each candidate nominated within an electoral district would be assigned a municipal unit. Parties were to allocate their seats to the candidates from the municipalities in which they had their strongest electoral performances. This was meant to provide an incentive for candidates to cultivate a local vote, thereby strengthening bonds between eventual legislators and their constituents. Party leaders, however, found innovative ways to circumvent the system and the plurality aspect did not have a serious impact on the allocation of seats.71 The innovative but ineffective system of seat allocation was replaced with a ‘flexible’ list system in 2004. Voters were provided with the option of casting a preference vote for a particular candidate on a party’s list. Candidates achieving a quota would receive a seat.72 In theory this provided voters a measure of control over their elected official. The quota, however, proved an unrealistically high bar for candidates to leap. 71 Manipulation occurred through several means. The electoral law did not clearly specify whether performance in a municipality would be determined by percentage of vote won or absolute number of votes attained. Electoral authorities allowed party leaders to choose the interpretation they most preferred. In addition, party leaders and electoral authorities agreed not to apply the malleable performance based criteria to seats that parties had acquired through the “largest remainder” round of seat allocation. Further requests by parties for post-hoc changes to seat allocation occurred at the discretion of the electoral authorities, particularly when it involved the allocation of seats to prominent party officials. An estimated 21% of seated legislators represented municipalities entirely different from that which they had originally been assigned (National Democratic Institute 1999). For more on the system and its application, see: (Crouch 2010, 49; D. Y. King 2003, 90-91; O’Rourke 2002, 199-200). 72 Quota being total votes in a district divided by total seats. 57 In 2008, The Law on General Elections also lowered the percentage of preference votes required to secure a seat, from 100% of quota to 30%. In the lead up to election, several large parties saw advantage in a pledge to allocate their seats based purely on preference votes. Candidates took the issue of seat allocation to the Constitutional Court, which ruled that all seats were to be distributed to those candidates receiving the most votes. As a result of the Court’s decision, Indonesia’s 2009 election would be an open list contest, with co-partisans competing for intra-party preference votes. In sum: since the fall of Suharto, institutional designers in Indonesia have made a series of electoral system reforms designed to prevent the emergence of regional parties, reduce the number of partisan competitors, and improve legislator-constituent ties. Whether these reforms were enacted sincerely for the public good or cynically for partisan gain is less important to the dissertation than simply establishing the empirical trends. These reforms have had a direct impact party organization and electoral campaigns. Indonesian party system: past and present The partisan actors After close to 30 years of stability under Suharto, Indonesia’s party system has evolved considerably in the Reformasi era. Most notably, the party system has experienced an expansion in the number of actors. This is not entirely clear simply by looking at the raw number of competitors. From 3 parties in 1997 (the last New Order election), the number of registered parties with ballot access jumped to 48 in 1999. It shrunk by half, to 24, in 2004. In 2009, there were 38 national competitors on the national ballot. The numerical ups and downs reflect a number of factors, from the explosion and waning of enthusiasm for the democratic process, the unsteady enforcement of electoral regulations, and the alterations to electoral laws.73 73 Given the stringent regional requirements, it is remarkable that so many parties achieve ballot access. In most cases, new parties are founded by activists who had previously worked together in an organization with a national scope. Pre-existing networks provide party founders with the local connections required to set up party offices in remote locations across the archipelago. Common organizational origins include pre- existing political parties, government institutions (most often the military), and religious bodies. Even parties formed to support the ambitions of an aspiring president typically have strong roots in one or more pre-existing national organizations. 58 The raw number of parties only tells part of the story however. One major noteworthy trend is that the major parties have become smaller over time. In 1999, the three largest parties won 68.8% of the vote; in 2009 they won 49.3%. The loss of votes for the largest parties produced an expansion in the medium-sized actors. Whereas commentators talked of the ‘Big-5’ in 1999, the category was expanded to the ‘Big-7’ in 2004 and by 2009 it made sense to distinguish between the 9 parties that crossed the legislative threshold from the 29 that did not.74 Describing all of the competitors in Indonesia is beyond the scope of this project. It is, however, useful to familiarize the reader with the major parties as discussion of these actors occurs frequently throughout the dissertation. Thankfully, there has been considerably continuity through time making is possible to precede chronologically. Table 3 – Major party performance % of National Electoral Vote Party 1999 2004 2009 PDI-P 33.7 18.5 14.0 Golkar 22.4 21.6 14.4 PKB 12.6 10.6 4.9 PPP 10.7 8.2 5.3 PAN 7.1 6.4 6.0 PKS [1.4]* 7.3 7.9 PD 7.5 20.8 Gerindra 4.5 Hanura 3.8 Top 5 Parties 86.6 66.3 63.2 Top 7 Parties 89.9 80.0 73.5 Top 9 Parties 91.5 85.1 81.7 *In 1999, Partai Keadilan Sejahtera ran as Partai Keadilan. In 1999, the ‘Big-5’ consisted of three authoritarian era parties and two new actors that sprung up to represent major religious organizations. With 34% of the vote, PDI-P was the clear winner of 1999. PDI-P is an offshoot of the New Order’s Indonesian Democratic Party (Partai Demokrasi Indonesia, PDI) and the heir of Sukarno’s PNI. Led by Sukarno’s daughter Megawati Sukarnoputri, the party’s voter base remains similar to the old PNI: secular Javanese and religious minorities. 74 The cut-off point is subjective and varies by author. Inclusion in the major party discourse at least in part reflects ex ante performance expectations of the pollsters. 59 Golkar, the incumbent party, managed to pick up 22% of the vote in 1999. Golkar had been Suharto’s party of hegemonic control. Throughout the New Order it was dominated by military personnel and bureaucrats. In 1999 it was particularly strong in Eastern Indonesia, where the bureaucracy played a particularly large role in the economic life of the citizens. Though officially a secular nationalist party, the National Awakening Party (Partai Kebangkitan Bangsa, PKB) was originally built on the back of the religious organization Nahdlatul Ulama. The party was closely associated with its charismatic and mercurial leader Abdurrahman Wahid, who managed to have himself selected president in 1999. PKB picked up 13% of the vote nationally. In 1999, PKB’s voting base looked remarkably similar to that of the Nahdlatul Ulama (Revival of Islamic Scholars, NU) party of the 1950s, with strong support coming from traditionalist Javanese and Madurese. The United Development Party (Partai Persatuan Pembangunan, PPP) earned 11% of the vote. PPP had been the New Order’s officially sanctioned Islam-oriented party. Hamzah Haz, the leader of the party in 1999, was an NU affiliated Malay from West Kalimantan. Haz’s eclectic background reflected PPP’s diverse religious and ethnic base. Despite the ability to appeal to a wide range of Muslim groups, PPP tended to perform strongest in the non-Javanese, modernist leaning areas. Like PKB, the National Mandate Party (Partai Amanat Nasional, PAN) is an officially nationalist party that relies on an established religious movement for organizational strength, in this case Muhammadiya. PAN’s leader in 1999, Amien Rais, was the former head of Muhammadiya and a prominent figure in the anti-Suharto protests. Despite Rais’s presidential aspirations, PAN only managed to win 7% of the vote. The party tended to do well among educated Muslims in urban centres. The 2004 election saw the ‘Big-5’ expand to the ‘Big-7’. The Democratic Party (Partai Demokrat, PD), formed only two years before the election, garnered 7% of the vote in 2004. The success of the ideologically non-offensive party is almost entirely attributable to its close association with popular presidential candidate Susilo Bambang Yudhoyono (SBY). Requiring a legislative faction to endorse his eventual presidential run, the popular former general encouraged his allies to form a party that could support 60 his ambitions.75 PD achieved its largest breakthroughs in urban centres, suggesting a middle-class basis. The Prosperous Justice Party (Partai Keadilan Sejahtera, PKS) was the other breakthrough success story of 2004. PKS had competed in 1999 under the name Justice Party (Party Keadilan, PK) but was forced to change names and re-register due to poor performance (less than 2% of the electoral vote). In 2004, the party gained 7% of the vote. Although the party was formed by Islamist students, the party’s success in 2004 was attributed to its strong anti-corruption message. Similar to PD, PKS had its largest breakthroughs in Java-based urban centres. Two more new players were added in 2009. Both were similar in form to SBY’s PD in that they were both founded to support presidential runs. The Greater Indonesia Movement Party (Partai Gerakan Indonesia Raya, Gerindra) was founded to support the political aspirations of Prabowo Subianto. A former special forces general and son-in-law to Suharto, Prabowo funded the party with his wealthy brother’s money. Gerindra promotes itself as a defender of the agricultural sector and a critic of ‘neo-lib’ economic policies. In 2009, the party gained 5% of the electoral vote. Support for the party was remarkably consistent across the country. The other new contender was the People's Conscience Party (Partai Hati Nurani Rakyat, Partai Hanura). Hanura formed to support the presidential ambitions of former general Wiranto. Wiranto had previously run as the presidential candidate for Golkar in 2004 but was disappointed in the lackluster support he received during his campaign. Determined to build his own brand, Wiranto ditched Golkar and established Hanura. The party finished with 4% of the electoral vote, no doubt a disappointment but still high enough to allow the party to seat its legislators. While new parties have managed to establish themselves as key players, no major party from 1999 has descended into the relative obscurity of the minor gurem (chicken- flea) parties. PPP, PAN, and PKB limp on, supported in part by their access to cabinet posts. Golkar and PDI-P have both experienced multiple party fractures which have eaten 75 SBY also played a behind the scenes role in the formation PPDK in 2002, though the leader of the party eventually severed relations due to SBY’s non-committal attitude (Mietzner 2009, 238). 61 away at their support bases, yet both still remain large by Indonesian standards. There has, in short, been remarkable consistency to the evolution Indonesian party system. In focusing on the major players, this description of the Indonesian parties has followed the conventional wisdom of party system change. Lost in the discussion is the remarkable expansion of the gurem vote. In 2009, the parties with no legislative representation attracted a percentage of the electoral vote (18.3), comparable to that of the country’s largest party. Though sometimes these parties have distinct ideological niches, the expansion of the gurem vote is best understood in the context of local dynamics. This dissertation will argue that gurem success has been a key factor behind party system expansion. So while the gurem typically attain minimal attention, they will become important players later in the dissertation. Parties and cleavage structure Aliran Now? The most common interpretation of the Indonesian political cleavages traces the divisions between parties back to broad societal divisions referred to as aliran, or streams. In its most basic form, aliran refers to the distinction between the devout Muslim santri and the religiously eclectic abangan. The two categories can be specified further, with the santri divided between traditionalist and modernist approaches to Islam and the abangan divided between those adhering to Javanese folk religion and those aristocratic priyayi elements which maintain a Hindu-Buddhist worldview. Each aliran group has its own distinct set of social and political organizations. Thus in 1955, each of the four large parties could be linked to a major socio-cultural grouping known as aliran.76 Several analysts highlight the similarity between the 1955 and 1999 election (King 2002, Lanti 2004). The comparison is particularly apt in the Javanese heartland. PDI-P, led by Sukarno’s own daughter, picked up support in former PNI areas. PKB, built on the foundations of the Nahdlatul Ulama religious organization and led by the organizations former head, fared well in those areas the NU party scored victories in 1955. Additionally, Amien Rais, the former head of the modernist organization 76 The four major parties and corresponding aliran groups were: 1) Priyayi: PNI; 2) Abangan: PKI; 3) Modernist santri: Masyumi; 3) Traditionalist santri: NU. 62 Muhammadiya, launched a party that did well amongst modernist voters. To many it seemed as though politik aliran had returned. Yet considerable differences exist. The intense conflicts between abangan and santri that characterized politics in the 1950s gave way to a more moderate form of competition at the turn of the 21st century.77 First, the election did not produce a clear successor to Masyumi. Devout Muslims in the Outer Islands supported a range of parties with a variety of positions on religious issues. Second, Intra-abangan competition could not easily be mapped onto the system. In particular, the existence Golkar, with its vague development ideology, appealed to both religious and non-religious voters and did not fit neatly into an aliran based reading of Indonesian politics. The party system de-aliranzation continued in 2004 and 2009. Three of the major players to emerge in this period – PD, Gerindra, and Hanura – can best be described as ‘secular nationalist’ parties that can appeal to devout and non-devout voters (Mujani & Liddle 2010, 36).78 As Tanuwidjaja (2010) points out, nationalist and Muslim parties have consciously tracked to the centre of the religious spectrum. For example, PDI-P formed a Muslim sub-organization and many of the major Muslim parties (PKS, PKB, PAN) have made efforts to recruit non-Muslims. The de-aliranization of Indonesian politics has, in part, been the consequence of institutional change. Liddle and Mujani (2007) find that voter leadership preferences drove partisan choices in all three post-Suharto elections. The leadership factor was no doubt enhanced by the shift to direct presidential elections, which has increased the incentive to run a broad-based legislative campaign. The three electorally successful parties formed after 1999 (PD, Gerindra, and Hanura) were all launched to support a candidate’s presidential ambitions (Yudhoyono, Prabowo, and Wiranto, respectively). In terms of policy positioning, little difference exists between the new presidential vehicles 77 Tracing the societal decline of aliran division is beyond the boundaries of the project. Two major factors are the general rise of personal piety and the triumph of constitutional secularism, both of which had their roots in the New Order. Suharto’s regime used religious education as an inoculation against communism, but the form of Islam that was propagated was private and conducive to the nationalist framework. At the time of transition the median voter was both pious and unenthused by calls the Islamic State. For more on santrification and de-aliranization, see: (Tanuwidjaja 2010; Ufen 2008). 78 Tanuwidjaja (2010, 32) prefers ‘secular inclusive’ for PD, a label that indicates the party’s open attitude toward religious voters. 63 and Golkar. The parties are designed to be non-offensive structures to support the politics of a broadly appealing president. But the effect of presidentialism is not limited to the rise of just these three parties. Less successful presidential aspirants, including Rachmawati Soekarnoputri, Lt. General M Yasin, Dr. Sjahrir, and Suharto’s daughter Siti Hardiyanti Rukmana (aka “Tutut”), all launched parties to support their presidential ambitions.79 In the case of the United National Democracy Party (Partai Persatuan Demokrasi Kebangsaan, PPDK), the founders began the party with the implicit understanding that the SBY, then an aspiring president, would eventually endorse it as his presidential vehicle. In other cases, parties formed with hopes of winning the support of a well known potential presidential candidate. For instance, the Prosperous Indonesia Party (Partai Indonesia Sejahtera, PIS) proclaimed their support for former Jakarta Governor Sutiyoso, while the Archipelago Republic Party (Partai Republik Nusantara, PRN) courted a public endorsement from the Sultan of Yogyakarta.80 The knock-on effect of party system presidentialization has been the reduced saliency of socio-economic differences. Two other factors that have increased the importance of personality relative to platform include electoral system change and direct executive elections at the local level. Electoral system reforms designed to tighten legislator-constituent linkages have increased the importance of individual candidates in the legislative campaign. Partisan symbols have taken a back seat to candidate posters and party machines have been replaced by the candidate-centered ‘team success.’ At the sub-national level, direct executive elections hastened the erosion of partisan brands. Ambitious politicians now regularly hop between party labels. Though de-aliranization has occurred, the cleavage has not been entirely eliminated. The residue of aliran remains, both in party units and voting behaviour. A 79 Respectively, the Pioneers Party (Partai Pelopor, PP), Functional Party of Struggle (Partai Karya Perjuangan, PKP), the New Indonesia Alliance Party (Partai Perhimpunan Indonesia Baru, PPIB), and the Concern for the Nation Function Party (Partai Karya Peduli Bangsa, PKPB) Arguably this is not a complete list. In some cases it is difficult to determine if the party’s principal patron held presidential ambitions before starting the party. For instance, Adi Sasono’s presidential ambitions were known but not publically proclaimed before he formed the ideologically non-offensive Freedom Party (Partai Merdeka). Eros Djarot, whose power struggle with Megawati prompted him to leave PDI-P, may have launched the Freedom Bull National Party simply to support his eventual presidential candidacy. 80 In both cases, party posters frequently carried the image of the candidate they planned to support, despite the lack of endorsement offered by the potential candidate himself. 64 traditionalist santri looking to vote for a religious party is still most likely to choose a traditionalist santri party over a modernist party, while a similarly devout modernist voters are more likely to choose a modernist option. The fact that these religious options exist and that voters in the same aliran family are more likely to circulate within a predictable range remains important. Still, the explanatory power of the aliran lens has declined. It cannot account for recent party system changes, except in the sense that the decline of aliran helps explain some of the free-wheeling electoral behaviour since 1999. Party Politics and Ethnicity Indonesia’s contemporary party system is not ‘ethnic’ in that party platforms and campaigns do not exclude sections of the electorate on the basis of suku bangsa. This has not always been the case. In 1955, the ethnically based Dayak Party (Partai Dayak) did well in West and Central Kalimantan. Likewise, the Awakening of the Simalungun People of East Sumatra (Kebangunan Rakyat Simalungun Sumatera Timur, KRSST) represented suku Simalangun in North Sumatra. Lax registration rules in 1999 allowed a few parties to participate that could be classified as ‘ethnic.’ The Unity in Diversity Party (Partai Bhinneka Tunggal Ika Indonesia, PBI) was widely viewed as a Chinese political party while the Father of the Orphans Party (Partai Abulyatama, PAY) was organized by Acehnese activists. Enforcement of strict regional organization requirements has since eliminated these borderline ethnic parties at the national and sub-national levels.81 Even though explicitly ethnic parties do not exist, ethnicity does directly affect voter behaviour. One mechanism involves the support for parties whose leadership candidates are viewed as co-ethnics. Thus Balinese identification with Megawati led to PDI-P support, Habibie lifted Partai Golkar among the Buginese, and Yusril Ihza Mahendra’s appeal among Melayu Belitung boosted the Crescent Star Party (Partai 81 The one exception to this rule has been Aceh, where in 2005 separatist rebels and national officials signed the Helsinki Agreement which provided the province of Aceh with an expanded form of political autonomy. Local parties were permitted to organize and compete for sub-national offices provided they meet proscribed organizational thresholds. In 2009, six local parties competed for sub-national legislative seats. All six local parties contained the term “Aceh” in their official name and the use of ethnic symbolism was prominent in campaign material. For more on Aceh’s political regulations and parties, see: (Barter 2011; International Crisis Group 2008). 65 Bulan Bintang, PBB) in Belitung.82 Preference for native sons (putra daerah) in leadership positions does exist, but it can be idiosyncratic. For example, it is hard to say the leadership of Batak Akbar Tandjung had any effect on Golkar’s fortunes in his home area of North Sumatra.83 As well, the limited number of visible leadership positions means only a few direct connections will exist between the party leader and a specific group or region. A pattern of voting for co-ethnic party leaders exists in a few cases, but the weight of the effect is hardly enough to determine large scale party system dynamics. Suryadinata et al. (2004) present the split between Javanese and non-Javanese as societal cleavage strongly shaping party preferences. This reading of Indonesian politics can be traced back to 1955, when three of the four major parties drew their votes largely from Javanese voter base (PNI, PKI, and NU). The same can be said of two of the successor parties (PDI-P and PKB) in the Reformasi era. Yet despite roots in the Javanese heartland, PDI-P and PKB do not promote themselves as exclusive parties and have often found success in non-Javanese areas. PDI-P, for instance, performs particularly well among non-Muslim ethnic minorities. Some of PKB’s strongest support has come from areas dominated by ethnic Madurese and the party has demonstrated its capacity to compete in non-Javanese areas.84 Parties vary in their reliance on Javanese voters but no major party is exclusive to that voting bloc.85 The non-ethnic orientation of Indonesia’s political parties at the national level may not be replicated sub-nationally. Liddle (1970) uncovered the ethnic bases of sub- national party competition during the parliamentary era. In multi-ethnic contexts of 82 Habibie is Buginese, and Mahendra is a native of Belitung. Though widely perceived as Javanese, Megawati’s Balinese grandmother provides a link to the island. This tenuous connection was no doubt strengthened by Islamist attacks labelling Megawati a Hindu. 83 Tandjung’s case shows the connection between leader and co-ethnics depends partially on self- presentation. Tandjung, married to a Javanese woman, tended to ally with Javanese politicians thrived in the world of Javanese politics (Tomsa 2006, 4). His ethnic ambiguity goes some way to explaining why he did not inspire the same type of co-ethnic loyalty engendered by rival leaders. 84 NU associated parties have a tradition of performing well in ethnic Banajerse areas, with PKB picking up at least one national seat in South Kalimantan in each election since 1999. In 2009 the party’s fifth highest vote total came from Maluku, a province with scarce few Javanese voters. In that case, the conditions for the PKB surge were put in place when the party nominated the wife of a prominent local official. 85 The same could be said of leadership positions of the parties mentioned above. PKB’s leadership tends to come from East Java’s eastern coast (‘tapal kuda’), a mixed Javanese and Madurese areas, which makes it multi-ethnic though not regionally representative. Megawati is considered Javanese, though her PDI-P has consciously included non-Javanese from the Outer Islands, like former Secretary General Alexander Litaay, in strong leadership positions. In addition, Palembang born Taufik Kiemas, Megawati’s husband, serves as both the head of the MPR and the party’s chief deal-makers, 66 Simalungun and Pematang Siantar, parties had distinct ethnic support groups: North Tapanulli Batak supported Parkindo, South Tapanulli Batak Masyumi, while the Javanese tended to support either PNI or PKI. Relying on Liddle’s description, Horowitz categorizes Indonesia’s party system in the 1950s as non-ethnic nationally but ethnic at the local level (Horowitz 1985, 301-2). In later chapters I will rigorously examine the hypothesis that local party systems simply reflect local ethnic competition. Gotong Royong and the power sharing tradition Discussion of party systems thus far has focused on particular actors and cleavage structures. Another important element of party system structure involves the interaction between parties in the legislature and cabinet. While parties compete at the polls, their relationships in the legislature are cordial to the point of collusion. Indonesian cabinets are routinely over-sized. Even parties ostensibly in the opposition maintain friendly relations with the government, especially when there are spoils to be divided. In this sub- section I trace the historical practice of Indonesia’s power-sharing tradition. Post-Independence Indonesia contained elements of both political polarization and power sharing. At the national level, cabinet coalitions contained ministers drawn from numerous parties, though important exclusions always existed. Parliamentary politics had its duelling poles in Masyumi and PNI, both of which led cabinets that excluded one another. It also had its cordon sanitaire in the form of PKI exclusion from cabinet. Cabinet politics, in other words, had winners and losers. The same could not be said of sub-national politics. Coalitions at the local level did not reflect always reflect national level cleavages and consensus based decision- making was the prevalent norm. Legge (1961) forwards two factors explaining sub- national cooperation. First, it accorded with deeply held norms of deliberation, consensus and mutual help. Second, sub-national politics were weighted toward the provision of local patronage, thus making it easier to cooperate with national-level rivals. Thus Legge found that Indonesia’s large sub-national coalitions had both a cultural and an instrumental logic. Sukarno desired power sharing at the national level. In his view, “50%+1” democracy was a foreign invention unsuited for the Indonesian context. To justify his 67 power sharing vision, Sukarno connected over-sized cabinets to cultural norms of mutual assistance: Let us form a Gotong Royong Cabinet. I expressly use the term gotong royong because this is an authentic Indonesian term which provides us with the purest likeness of the Indonesian spirit. The cabinet should include all political parties and groups represented in parliament which have obtained a certain quotient of votes in the election. (Feith and Castles 1970, 82-385) The initial effort was opposed by the religious parties. Masyumi and, to a lesser extent, NU refused to sit in government with the PKI, a party they viewed as atheist. Sukarno, however, saw the political tides were turning against the parties. He introduced the ‘Nasakom’ concept, a proposal for coalition encompassing the major nationalist (PNI), religious (NU) and communist (PKI) parties.86 Though Sukarno never managed to win over a sceptical Army on the merits of PKI participation, his effort to bridge seemingly incongruent ideologies was a forerunner of coalition politics in the Reformasi era. Sukarno liked the symbolism of power sharing more than he liked the practice. His preference for power sharing symbolism did not rub off on Suharto, though his aversion to opposition did. During the New Order, cabinet posts were assigned only to members of the Golkar family. Despite his exclusion of PPP and PDI from cabinet, Suharto made it clear the legislature was not to function in terms of a government- opposition dynamic (Elson 2001, 189). Non-Golkar parties occasionally voiced criticism of specific efforts and the voting public came to treat them as opposition parties, but they tended to cooperate within the legislature.87 The regime rewarded them with subsides to continue their operations. The transition to democracy ushered in a new period of power sharing politics. Wahid’s initial “National Unity” cabinet contained representatives from all the major parties and many of the minor ones. When he alienated his partners by dismissing politically sensitive cabinet ministers, the broad coalition that put him in power replaced him with Megawati. Megawati formed a Gotong Royong cabinet that included members 86 ‘Nasakom’ is an acronym for the ideological streams Sukarno was trying to include (Nasionalisme, Agama, Komunisme). 87 Legislators from the armed forces claimed to be more willing to challenge state policy than the opposition parties (Ziegenhaim 2008, 51). 68 from the five major parties, including politicians from PKB, the party of recently ousted Wahid. After SBY replaced Megawati in 2004, he brought all major parties but the PDI-P into his United Indonesia cabinet. Following the 2009 election the list of excluded parties grew to include Gerindra and Hanura. Yet the growth of excluded parties does not indicate a polarizing political climate. Members of the ‘opposition’ frequently find themselves in negotiation about cabinet entrance and the government maintains cordial relations through the distribution of important legislative posts.88 Signs of a similar power sharing dynamic can be observed at the local level. Local legislatures are frequently characterized by broad cross-party coalitions.89 Party alliances created for the purpose of electing municipal leaders rarely follow an ideological logic; the Islamist PKS can be found in alliances with the secularist PDI-P and the Christian Prosperous Peace Party (Partai Damai Sejahtera, PDS). Executive elections do provide a level of conflict and a distinction between winners and losers. Yet when it comes to skimming the budget coalitions are not simply ideologically incoherent but near universal in scope. In sum: Each election has witnessed the rise of new medium sized players, though the total vote collected by the large parties has declined over time. The importance of Indonesia’s traditional aliran cleavage structure has declined over time. While the existing parties have a non-ethnic basis, there is evidence of both long-term and leadership driven patterns of ethnic voting. Lastly, Indonesia’s power sharing tradition has re-emerged in the reformasi era. Rent opportunities and decentralization A central claim of the dissertation is that local rent opportunities affect party system outcomes. Potential access to local spoils influence the likelihood an elite will enter the political realm, the choice of party affiliation, and the method of campaigning. But this raises several questions: Why are local rents important in Indonesia? How can they be 88 A prime example was the installation of prominent PDI-P figure Taufik Kiemas to the position of Speaker of the MPR. The governing coalition, with a large majority in the national legislature, could have easily held the post if it was inclined to do so. 89 On sub-national dynamics, see: (Slater 2004. 63; Ufen 2008, 31). 69 measured? And why do they correlate with ethnic diversity? Before exploring rent opportunities, however, the more immediate questions of why and how Indonesia decentralized authority to sub-national units must be addresses. Decentralizing Indonesia ‘The big bang’ Independent Indonesia has struggled to find the right balance between the centre and regional governments. In the 1950s Indonesia experimented with decentralization program and, much to the chagrin of the centre, occasionally transferred de-facto authority to regionalist movements. Overall, though, the trend during the Sukarno era was toward greater centralization, in law if not always in practice. The centralizing tendencies of Sukarno’s Guided Democracy regime were entrenched during Suharto’s New Order, with early economic reforms strengthening the fiscal powers of the central authorities (Anderson 1983, 488-490). Whereas corruption during the post-independence period had been party organized (outside of the military), frequently decentralized, and often chaotic, Suharto sought to centralize and streamline state corruption. A complex pyramid of corruption was created, and Suharto sat at the very top. Governance was accomplished through payoffs and kick-backs, with each level of the pyramid making sure to payoff their superiors. The patronage structure ensured that a large group of administrators were loyal to Suharto and that the business of corruption was conducted in an orderly manner. The fall of Suharto opened the door to the possibility of significant change in centre-periphery relations. During the period surrounding Suharto’s fall numerous regions erupted in violence as societal forces jostled for power in a new system. In the capital there was fear that the country was coming apart at the seams. Decentralization was grasped as a tool to fulfill a demand for change in the regions and head off separatist sentiment. It also happened to be part of the broader international zeitgeist of ‘good governance’ that prevailed in the late 1990s and international organizations were involved in discussions with the crafters of the decentralization law.90 90 This being said, the group tasked with writing the law (‘Tim tujuh’ or ‘Team of Seven’) were committed to decentralization independent of any prodding by the international donor community (Turner et al. 2003, 5-6). 70 Laws 22/1999 and 25/1999 constitute the two fundamental pieces of the reform package, which Turner and Podger refer to as “the most radical decentralisation measures in Asia and the Pacific” (Turner et al. 2003, xii). Law 22 sorts out the power relationships between the municipal, provincial, and national levels of government. In an unusual move, the authors of Indonesia’s laws bypassed the provincial level governments and transferred to municipalities wide powers over all but five areas specifically reserved for the national level.91 In transferring power to municipalities rather than provinces the central government hoped to gain the support of local leaders without empowering provincial administrations that may quietly yearn for separation. Whereas municipalities were previously sub-ordinate to provinces, the vertical chain of command was reformed and provinces became mere assistants to the municipal and national governments. Importantly, sub-national units were given the right to select their own leaders free of interference from higher levels.92 Law 25 gave sub-national units the right to expand their own tax base and keep a greater share of the revenue from natural resource extraction. Funding of municipalities was to be done through transfer payments, known as General Allocation Funds (Dana Alokasi Umum, or DAU) and Special Allocation Funds (Dana Alokasi Khasus, or DAK).93 In effect, the national government would collect and disburse funds to the local government with few strings attached.94 The extent of the decentralization program and its occasionally messy implementation caused concern in Jakarta. An additional series of reforms in 2004 curbed a few of the perceived excesses. Mirroring reforms at the national level, municipal executives were given their own elected mandate and the legislature’s right to dismiss the 91 These included foreign affairs, defence and security, justice, religion, and monetary/fiscal affairs. 92 During the New Order, the central government played the leading role in the selection of provincial governors, while the governors were the key actors in the selection municipal executives. 93 DAU constitute the largest component of direct transfers (approximately 85-90%). The central government is obligated to disburse at least 25% of its annual revenue through DAU transfers, which are allocated to sub-national units based on a formulas taking into account population size, geography, construction prices, and other factors. DAK transfers, on the other hand, are targeted at selective project and state priorities. 94 Some municipalities have access to significant internal sources of natural resource revenue. The decentralization law have these resource rich municipalities the right to keep a larger portion of their earnings than they previously had. Most municipalities, however, are dependent upon transfers from the central government. 71 executive was substantially curbed. The authorities of both the central and provincial governments were strengthened at the expense of the municipalities. Nonetheless, the financial and legal position of municipal governments remained strong. Decentralizing Corruption Decentralization began with the hope that local accountability would improve governance processes and outcomes. Most observers, however, find that the decentralization of authority was accompanied by the decentralization of corruption (Crouch 2010, 111; Hadiz 2003, 1222; van Klinken and Nordholt 2007, 18; Korte 2011, 22; Malley 2003, 115; Ufen 2008, 32). Whereas the New Order structure of informal politics was top-down and centralized, the transfer of power to sub-national units and the democratization of the regime created a more diffuse pattern of corruption. Hadiz provides a succinct description: While the centralised system no longer exists, its elements have been able to reconstitute themselves in new, more fluid, decentralised and competing networks of patronage. The range of interests now competing at the local level are even more varied than under the New Order. They include ambitious political fixers and entrepreneurs, wily and still-predatory state bureaucrats, and aspiring and newly ascendant business groups, as well as a wide range of political gangsters, thugs, and civilian militia (Hadiz 2003, 124) While decentralization and the fall of Suharto shook the old power structure, New Order elites did not disappear. Rather, they were supplemented with new groups of elites seeking to control valuable offices. To return to the quotable Hadiz, “[A]s in the Philippines following the fall of Marcos, the driving logic of political life in Indonesia remains the ‘quest for rent-seeking’ opportunities through the securing of ‘access to the state apparatus’ for the purposes of private accumulation” (Hadiz 2003, 122) . Not only have sub-national patronage networks fragmented, the substantially increased funds flowing into the municipal governments augments the pay-offs available to local actors. As Shulte Nordholt and van Klinken put it, “Because most regions are subsidized by the centre, regional governments tend to become spending machines” (2007, 17). These ‘transfer rents’ frequently find their way into the private accounts of local political elite. Expanded authority to raise revenue also provides the opportunity to craft local regulations and levy taxes in order to protect friendly businesses and/or increase the local budget for the purpose of looting. Likewise, the transfer of a 2.44 72 million civil servants from central to regional government control provided new opportunities for extortion and patronage. Nothing quite signifies the bonanza of opportunities better than the expansion of new sub-national units. The practice of municipal and provincial creation has taken the name pemekaran, which quite literally means ‘blossoming.’ Between 1998 and 2009 there were 185 new municipalities, an increase of 63%. Added to that, 7 new provinces were created. The over-riding motivation behind pemekaran has been the expanding access to state resources. Each new sub-national unit means buildings that need to be built and offices that need to be filled. Given that revenue comes in the form of transfer payments, the cost of a new municipality is largely borne by the national government. The revenue allocation formulas that determine transfer payments are written in a way that, in the aggregate, a municipality will almost always enjoy a net benefit from splitting into two distinct units. Thus in many cases the drive for new municipalities appears simply as a method of achieving increased revenue from the centre. The mechanisms of local corruption Sub-national political competition revolves around accessing state resources. When it comes to abusing office for personal and political gain, Indonesian legislators have an impressive menu of manipulation. Legislative sessions begin with a vigorous round of cow-trading in which the various factions jockey to fill the most lucrative committee and legislative posts. Committee members then use their over-sight function to extract resources. Businessmen, bureaucrats, and members of the executive provide gifts and favours to secure favourable policies, contracts, or simply autonomy. Many legislators take the opportunity to enter the world of contracting, funnelling government money to projects benefiting their own companies. Even the legislator not sitting on a particularly ‘wet’ (basah) committee can still find ways to profit. The most straight-forward method involves simply increasing their financial compensation package. Legislators routinely collude when it comes to legislative salaries and kickbacks. In West Java, all 100 sitting members were investigated for granting themselves a large and potentially illegal increase in their living allowance (Crouch 2010, 240). In West Sumatra, the vast majority of provincial legislators were successfully prosecuted with illegally raising their personal salaries 73 (Davidson 2007). Likewise, all 44 members of the West Papua legislature were named as suspects in a scheme that found legislators accepting illegal cash transfers from a publically owned company (Tempo 2011). It is now common to find the informal existence of universal coalitions aimed at transferring state resources to sitting legislators. Merely holding a seat guarantees legislators a healthy slice of the pie. Additionally, the legislative process provides plenty of opportunities to cash in on political power. Support for a piece of legislation or a political appointment can be attained through cash payments. Before direct elections were introduced, the executive routinely had to pay legislators in order to avoid rejection of the accountability speech, an event tantamount to a no confidence vote. The opportunity to sell support has declined since direct executive elections, though chances to shake-down an executive branch continue to exist. Executive offices are even more lucrative than legislative posts. Municipal heads take the lead in appointing the top civil servants and department heads in the region. These tend to be close allies or individuals with useful political connections. Top government officials use posts they control to build patronage networks and extract payments from aspiring bureaucrats. Before a regional head can access state resources, however, she must first win election. Prior to the 2004 reforms, executive contests involved legislative coalition building, typically facilitated through direct cash payment to legislators. Since direct elections, an executive team requires support from a party or coalition with 15% of the electoral vote or 15% of the legislative seats. Local party leaders capitalize on the law by selling their support to a needy executive ticket. Given that direct elections for sub- national offices are primarily personality contests in which programs and party identification play little role, coalitions of convenience are common. Aspiring executive tickets pay large sums of money to “rent a boat”, a practice that involves candidates cobbling together a coalition in order to ensure ballot access. Candidates will sometimes have no prior connection to the parties supporting them in the coalition. In the municipality of Gowa, for instance, the former Golkar activist Hasbullah Djabar bought 74 the support of PAN and Partai Merdeka, allowing him access to the ballot.95 Interestingly, even parties that do not attain a seat in the municipal legislature still act as important coalition players. Sjachrir Sjafruddin, also a Golkar stalwart, cobbled together a coalition of 13 parties to attain ballot access in Gowa, of which only PKS had a local representative. In sum: Contemporary Indonesian politics is driven by rent-seeking. Due to power-sharing dynamics, even legislators from minor parties can enrich themselves and their associates. Yet not all legislators throughout the country face the same opportunities to access rents. Rent opportunities and ethnic diversity A politician’s opportunity to engage in rent-seeking behaviour is dependent upon the extent of state involvement in the economy and the constraints placed on legislator behaviours. Administrative districts with high rent-seeking opportunities combine a mixture of both high state involvement in the economy and weak constraints on legislator behaviour. In this sub-section I operationalize the rent opportunities concept using available data, demonstrating that the resources available to politicians and the constraints on their actions vary across administrative districts. Furthermore, variance in these key metrics tends to correlate with ethnic diversity.96 State Resources State jobs in Indonesia are a valuable commodity and serve as a rough measure of state involvement in the economy. Indonesians will sacrifice substantial money and effort to attain a position for either themselves or their kin. Despite low official pay rates, discretionary salary top-offs and bribe revenue ensure a reliable and comfortable wage for most state employees. Educated, upwardly mobile citizens may hope to secure employment as an administrator in one of the country’s many local bureaucracies; those with lesser formal education can hope to be hired as part of the vast army of maintenance 95 For a detailed description of the Gowa Municipal race see: (Buehler and Tan 2007) 96 This is not to deny that rent opportunities are affected by factors beyond ethnic diversity, including resource revenue. As the results from this section indicate, ethnic diversity is only a partial explanation for variance in sub-national rent opportunities, though the relationship is strong and politically consequential. 75 workers, drivers, security personnel, or other low-skill positions that service Indonesia’s state departments. Direct manipulation of hiring practices is considered the norm.97 Access to state jobs is part in parcel of the Indonesian politician’s political appeal. The expansion of state employment opportunities is also implicit in policy positioning. For instance, during the run-up to the 2009 legislative election in North Sumatra the issue of new administrative district creation captured considerable local attention. Politicians of various stripes championed the division of North Sumatra into two provinces. Not lost on onlookers was the implication of this division: new political offices, new departments, and new jobs.98 Indonesian politicians not only offer assistance securing employment, they actively work to expand their opportunities to deliver patronage in the future. The relative weight of civil service employment within the modern sector has consistently correlated with ethnic fractionalization scores for the last three decades.99 Figure 5 demonstrates the correlation in 2005. Ethnically homogenous provinces tend to have smaller bureaucracies. This finding is statistically significant both before the fall of Suharto and before every election held in the Reformasi era. Ethnically diverse administrative areas have a larger share of their modern sector workers employed in the civil service. 97 As the World Bank notes, the civil service is characterized by informal payments for entry and promotion (2007, 16). The informal market for jobs is partially structured by social and political relations. Applicants who are perceived as socially distant can be subject to a different bribery rate. Aragon has called this practice ‘unequal opportunity buying’ (2007, 41). Patronage networks are important as they increase the probability of favourable treatment and these networks often (but not always) follow ethnic or religious lines (Aragon 2007; van Klinken 2007). 98 This particular pemekran campaign ended tragically when a rent-a-mob hired by proponents of the new province frightened the provincial Speaker of the House to the point of a deadly heart-attack. The mob’s paymaster, a candidate for provincial office running under a minor party label, was arrested and tried for his role in the affair. 99 ‘Modern sector’ refers to non-agricultural work. My focus on the modern sector follows established practice both within the comparative literature and the Indonesia literature (Chandra 2004; van Klinken 2007). 76 Figure 5 – Ethnic diversity and civil service size Large bureaucracies in diverse areas are primarily funded through transfers from the centre.100 A simple examination of central government transfers by population and ethnic fractionalization reveals a statistically significant correlation between the two variables.101 Diverse areas receive relatively more money from the central government and can thus afford to hire relatively more bureaucrats. Taken together, the data on transfers and civil service size indicate that the magnitude of state involvement in the modern economy tends to be greater in ethnically diverse areas. Constraints on Behaviour Legal and normative constraints on legislator behaviour constitute the second aspect of the rent opportunities concept. The most direct measures available for the patterns of policy implementation are corruption perception indexes. The Transparency International Indonesian Corruption Index provides one measure of corruption perceptions in 50 cities across Indonesia (Transparency International Indonesia 2008, 2010). Using surveys of businesses leaders, TI has developed a 0 to 10 corruption score for Indonesian cities, with a 0 rating being “very corrupt” and a 10 rating being “very clean.” 100 See Appendix C, Section 2. 101 See Appendix C, Section 3. 77 Figure 6 plots the relationship of an aggregate corruption score and ethnic diversity.102 The X-axis represents a province’s ethnic fractionalization; the Y-axis captures average corruption scores within the province’s cities, with corruption increasing as TI scores decline. A clear relationship exists: cities in ethnically diverse provinces are perceived as more corrupt. The relationship is statistically significant, even controlling for key variables like average income and population size.103 Additionally, the relationship is stable when the scores are disaggregated by year; a positive relationship existed in both 2008 and 2010. Figure 6 – Ethnic diversity and corruption Where legal authorities and the electorate abide corrupt practices, public service delivery tends to suffer. First, resources can be re-routed from public goods provision to spending that is more amendable to direct manipulation. For example, funds diverted from road water treatment and distributed to agricultural aid or general administrative costs are easier for politicians to selectively distribute. Second, the delivery of an ostensibly public good can be perverted by patronage politics. A politician that secures university admission for political supporters or provides electrical lines to only supportive villages erodes the universality of a public good, be it education or the provision of electricity. 102 The corruption index averages the corruption score for 2008 and 2010, the two years data are available. 103 See Appendix C, Section 4. 78 Data on delivery of public services comes from the Indonesian NGO Regional Autonomy Watch (Komite Pemantuan Pelaksanaan Otonmi Daerah, KPPOD).104 Like the TI datasets, KPPOD measures service delivery by municipality. I use their Local Infrastructure Sub-Index as a metric of public service delivery. Using results elite surveys, KPPOD assigns each municipality in their sample a score based on their ability to delivery key infrastructure services. Quantitative tests reveal a strong correlation by ethnic diversity and low infrastructure scores.105 The finding is statistically significant in both 2003 and 2007 datasets (Komite Pemantuan Pelaksanaan Otonmi Daerah 2003, 2008). To demonstrate the strength of key relationship, Figure 7 plots 2007 KPPOD provincial averages by provincial ethnic fractionalization. The X-axis represents provincial ethnic fractionalization scores. The Y-axis captures average infrastructure scores within the province. For the 2007 sample, municipalities were assigned an infrastructure score ranging from 0 to 100, with infrastructure quality increasing as scores increase. There is a clear correlation between the variables, with average infrastructure scores decreasing as ethnic fractionalization increases. Not only do elites perceive diverse areas to be more corrupt, data also suggests that politicians and bureaucrats in these regions do not invest in public service delivery. 104 The organization works closely with USAID and the Asia Foundation to monitor the effect of decentralization on investment climate. 105 See Appendix C, Section 5. 79 Figure 7 – Ethnic diversity and infrastructure quality Diversity and rent opportunities: explaining the relationship A strong relationship exists between ethnic diversity and the various measures of rent opportunities. Diverse areas receive more transfers from the central government and their modern sectors have a higher proportion of civil servants, factors indicating substantial state involvement in the economy. Likewise, diverse areas are perceived as more corrupt and tend to under-provide public services as compared to homogenous areas, factors indicating weak legal and normative constraints on elite behaviour. There are two primary reasons the patterns occur. First, due to legacies of state building, diverse municipalities and provinces tend to be smaller in population. Second, ethnic diversity reduces the capacity to enforce anti-corruption norms. Legacies of State Building Ethnically diverse administrative units tend to have smaller populations. Geography and pre-colonial agriculture combined to generate this demographic fact.106 Yet it was by no 106 The roots of the correlation between ethnic homogeneity and unit population size can be found in Indonesia’s agricultural history. In the pre-colonial era, wet-rice cultivation enabled population growth. Those groups that combined agricultural techniques with a favourable geographic endowment experienced population expansion. The agricultural surplus also provided an economic base for strong kingdoms 50 60 70 80 Av era ge KP PO D S co re 0 .2 .4 .6 .8 1Ethnic Fractionalization Average KPPOD Scores by Ethnic Fractionalization (Provincial) 80 means inevitable that administrative units in diverse areas would have small populations. At different points of time the whole of Sumatra, the Indonesian portions of Kalimantan, and all the entire areas east of Kalimantan (excluding Papua) formed independent governance units roughly equivalent to a contemporary province. Large, diverse units were a possibility. The final outcomes of small, diverse units reflected a political process of state building that involved both coalition building and coercion. Early Indonesian state builders confronted a dauntingly complex structure of territorial governance. In the colonial era, Dutch authority had been spread using an array of contracts, ultimatums, and conquests.107 The legal authority of subjugated local rulers and the system of administrative oversight varied throughout the archipelago. Even with the rationalization that occurred during the post-war federal system, the Dutch still utilized 11 different types of local government that ranged from the familiar municipal governments and federal districts to more contextual “neo-lands”, “negaras”, and “neo- group communities” (Shiller 1955, 89-90). An early task of the Indonesian state was to bring coherence to the colonial patchwork. After various iterations, they arrived at a system of provinces and municipal units referred to as kabupaten and kota. This pattern was most familiar to Java, where the Dutch had set up a similar division of territory.108 The fit was awkward in the Outer Islands. Many smaller self-governing units were consolidated into artificial borders, though this outcome was likely given the dispersion of the population and the localization of identities. At the provincial level, the government created 10 provinces.109 Half of these were located on Java. The post-Independence territorial consolidation did not last. In a few cases, regional elites who had previously enjoyed a measure of autonomy resented their reduced status. In some cases, coalitions of local elites and military commanders led rebellions capable of reinforcing cultural unity. In contrast, areas less conducive to wet-rice tended to rely on swidden agriculture practices that enabled neither population growth nor land-based kingdoms. Consequentially, Indonesia has been left with a few densely populated pockets of ethnic homogeneity and many geographically large tracts of sparsely populated and ethnically diverse lands. 107 For a brief overview of the Dutch territorial structure, see Legge (Legge 1961, 21-6). For a comprehensive account, see (Schiller 1955, 80-103). 108 At one point these were simply “first-level” and “second-level” so as not to offend the Outer Islands population with Javanese terms (Legge 1961, 13). 109 This count included Yogyakarta and Jakarta, which were considered special regions with the status of a province. 81 against centralized control. These phenomena created momentum for new territorial units. In response, the central government adjusted boundaries in order to reward allies and isolate rebels. Acehnese rebels were offered a new province in 1956, assuaging anger at their forced incorporation into North Sumatra province.110 Minangkabau rebels in Central Sumatra were contained when the central government provided east coast elites with two new provinces (Riau and Jambi). In Kalimantan, Jakarta eased regional complaints and isolated rebellious Banjarese by splitting the territory into 4 provinces. These boundary adjustments and others not only provided symbolic benefits to the affected populations, they also allowed the central government to incorporate cooperative elites into the state structure while credibly committing to address their concerns in the future. By the early 1960s, all of the large, diverse provinces in the Outer Islands had been cut up into smaller units. The Suharto era was marked by the relative stability of administrative borders. As Suharto’s rule drew to a close in the late 1990s, regional discontent exploded throughout the archipelago. Again, the centre utilized administrative boundaries to assuage critics, isolate and contain restive groups, and frustrate potential separatists. In provinces with separatist movements there was a large expansion of new municipalities. New units meant new prominent positions which could be doled out to loyalists. This process helped solidify a pro-state constituency in contested areas like Aceh and Papua. Similarly, provinces struck by communal conflict saw a flurry of re-districting. In particular, Central Kalimantan, West Kalimantan, Central Sulawesi, and Maluku experienced considerable boundary change as the state sought to isolate rival ethnic groups. In some areas, such as North Sulawesi, boundary change pre-empted violent communal conflict before it could occur. All of these boundary changes were driven in part by the need to establish and maintain internal security. Though they did not exclusively affect diverse units, there was a strong tendency for diverse units to be split as they also tended to be the sites of conflict. 110 Rebel leader Daud Bureah continued to hold out for a more substantial autonomy package that included increased authority over the regulation of religion. 82 District creation was not confined to regions with security concerns. The easy process of re-districting contributed to an explosion of new demands.111 Most of the split districts were low-density units in the Outer Islands (Fitrani, Hofman, and Kaiser 2005, 76). There was also a tendency for the split municipalities to have large pre-existing civil service expenditures. This suggests the municipalities already reliant on state transfers were more likely to split, and most of these rent-seeking splits occurred in diverse units. Relatively peaceful but diverse provinces like Bengkulu, East Nusa Tenggara, and Southeast Sulawesi experienced significant fragmentation. Re-districting only increased the comparatively large transfers that flowed into these regions. Though re-districting has certainly produced a homogenization at the municipal level, splits were not limited to diverse units. In East Nusa Tenggara, ethnically homogenous municipalities like Manggarai and Sumba Barat were split and sometimes re-split.112 The end result has been that diverse regions have seen their administrative boundaries re-drawn in ways that have significantly decreased the size of the units themselves. Diversity and Corruption Long term state-building processes ensured that the state would play a significant role in the local economies of diverse areas in Indonesia. It is not simply the relationship between diversity and small population size that drives the correlation between ethnic diversity and rent opportunities however. Part of the explanation lies within the political dynamics of diverse societies. Specifically, diverse jurisdictions in Indonesia tend to be 111 The basic legal framework for the creation of new municipalities is found in Government Regulation 129/2000. New municipalities must be recommended by the legislature and executive of the parent district, the governor of the province, and the Ministry of Home Affairs (MoHA). These hurdles are largely political, however. New municipalities do not have to meet any stringent criteria. The main criteria listed in law are simply that a new district must contain a minimal number of municipal administrative units (Kecamatan), though this is largely a formality as Kecamatan creation is not an administratively difficult task. The MoHA is supposed to conduct a technical analysis of a proposed municipality and make its recommendation. This analysis is supposed to carefully consider a range of 19 indicators covering economic, socio-cultural, and socio-political concerns (USAID 2006, 19). Despite having the formal authority to conduct intensive reviews and analyses, the MoHA does not use this power in a systematic manner. Pro-forma reviews are common. Even if the MoHA were to apply increased scrutiny, the national executive branch can potentially be bypassed altogether if supporters of the new division can successfully lobby the national legislature to create the new municipality in lieu of executive support. 112 Sumba Barat did have relevant religious divisions between Protestants and Catholics that may have motivated the split. Closer examination of the most recent census may uncover that splitting in ethnically homogenous areas still had underlying communal motivations. 83 corrupt for two inter-related reasons: 1) the breakdown of anti-corruption norms; 2) zero- sum budgetary processes. The breakdown of anti-corruption norms in diverse political systems is especially common. Breakdown occurs in part because anti-corruption norms are subsumed by ethnic bias and/or a broader social stability norm. The threat of communal conflict tends to provide a backdrop to political discussion in ethnically diverse settings. If accused of corruption, corrupt actors often find it useful to portray their accusers as ethnically biased, a defence that is most compelling when the accuser is not a co-ethnic. Pre- emptive gifts to co-ethnics help shore up support and prevent particularly damaging accusations from emerging. When a corruption allegation becomes an ‘ethnic’ rather than a ‘governance’ issue, societal leaders feel pressure to place the entire problem in abeyance for the sake of social peace. The public, for its part, may discount accusations made against co-ethnics as bias. With the public divided and the elites weary of conflict, officials experience minimal backlash for engaging in corrupt behaviour. Davidson’s work details the mechanisms that explain the broader correlation between diversity and corruption revealed in the corruption perception data (Davidson 2007). In his examination of anti-corruption campaigns in West Sumatra and West Kalimantan, Davidson finds distinct ethnic dynamics hindered anti-corruption campaigns in diverse contexts. In homogenous West Sumatra, middle-class civil society activists built an organization that generated enough pressure to bring about charges on the large majority of the provincial legislature. The success of the efforts led to similar investigations throughout the province. In West Kalimantan, on the other hand, anti- corruption campaigns simply reflected the ethnic political jostling that defined the provinces politics. Specifically, allegations against the Dayak dominated executive in Mempawah were pressed by a largely Malay civil society organization. The Dayak executive mobilized ethnic supporters and managed to steer investigators towards their Malay legislative rivals. Unlike the strong anti-corruption organization that emerged in West Sumatra, the organization in West Kalimantan was fleeting as it was meant only to score short-term political advantage. Davidson notes, “the thickness of ethnic politics smothered its issue counterpart” (2007, 90). In short: the anti-corruption campaigns in 84 West Sumatra were about good governance, while those in West Kalimantan were about ethnic competition. Diverse governance units in Indonesia have also been characterized by zero-sum budgeting dynamics that enable (and perhaps encourage) corruption. The provision of universally accessible public goods is viewed suspiciously by those who suspect their ethnic rivals may benefit more than co-ethnics. The erosion of trust facilitates spending on particularistic goods that can be manipulated by corrupt actors. This argument is difficult to empirically pinpoint, but telling signs exist in the existing case literature. Aragon’s (2007, 51-60) description of governance in Poso reveals the tight relationship between ethnic competition and corrupted infrastructure projects(Aragon 2007, 51-60). Bandiera & Levy find that diverse villages that are governed democratically provide fewer social services that benefit the poor, such as hospital beds and low educational fees (Bandiera and Levy 2011).113 In his study of local governance, von Luebke (2009) simply assumes ethnic diversity will impact service provision and carefully controls his case-selection so as to only include homogenous municipalities. In their brief survey of inter-municipal cooperation, Turner & Podger find ethnic homogeneity provides policy-makers with a less-challenging environment in which to build lasting projects. Regarding a solid waste project in Bali, the only successful inter-regional project the authors could find, Turner & Podger note, “[The project’s] success was largely due to the social coherence of Balinese society at the subregional level, as well as pressure from the tourist industry” (Turner et al. 2003, 95, emphasis added). More commonly, authors have focused on the ethnic competition for top political and bureaucratic positions as it has simply been assumed that key posts provide valuable bailiwicks for specific groups (R. William Liddle 1970, 159-163; Tanasaldy 2007, 359-371). While not all of the literature mentioned focuses directly on the budgetary process, they all draw a link between ethnic competition and governance outcomes. An 113 Interestingly, Bandiera & Levy also find that diverse areas provide increased community security measures and more days of voluntary labour. The authors theorize that diverse systems empower the wealthy elite by allowing them to form a coalition with an ethnic group. Thus the over-provision of some public goods and the under-provision of others reflect the relative relative power of the elite, which prefer security programs over health and educational services. 85 underlying intuition exists that ethnic competition reinforces an Indonesian tendency to treat budgeting as a division of the spoils. This intuition is consistent with the strong, negative correlation between infrastructure provision and diversity. In sum: This section has demonstrated that rent-seeking opportunities correlate with ethnic diversity in the Indonesian context. First, diverse provinces have a greater percentage of modern sector workers employed in the civil service. These large sub- national bureaucracies are funded through transfer payments from the central government. Second, politicians in diverse sub-national systems face fewer constraints on their behaviour, as evidenced by high perceptions of corruption in diverse cities. Consistently low performance in infrastructure provision suggests politicians funnel money away from public services and towards more particularistic forms of state spending. In part, these outcomes are simply endogenous to diverse political systems. Competition between groups weakens anti-corruption efforts and incentivises particularistic spending to support co-ethnics. Nonetheless, some of the correlation between rent opportunities and diversity is a consequence of low population size in diverse areas. The state-building process in Indonesia has produced administrative fragmentation in diverse regions, increasing the reliance on state employment in these areas. Conclusion Indonesia’s aversion to a fragmented, ethnically divided party system has deep roots. Indonesia’s early nationalist leaders were deeply suspicious of ethnic and /or regional political organization. Their suspicion was reinforced by traumatic events, including the divide and rule tactics of the Dutch, the regional rebellions of the 1950s, and the recent violence that accompanied the downfall of Suharto. The aversion to sub-national political organization has been built into the country’s electoral institutions. Strict party registration and general elections laws ensure that sub-regional parties do not make it onto the ballot. These laws have been strengthened with each election. Likewise, the aversion to a fragmented party system stretches back to Sukarno’s attack on the parties in the 1950s. After each election, institutional reformers have tightened the screws on smaller parties. 86 The party system has not evolved in the way reformers have intended. Each election has witnessed the ascension of new modestly sized parties and, just as importantly, the survival of the established political players. The vote share for the small parties has risen to surprising heights. This fragmentation of the system is not indicative of any societal polarization however. The importance of the old aliran cleavage has faded with time. Existing parties have moved to the ideological centre to join the new, ideologically inoffensive presidential vehicles that were formed after the introduction of direct executive elections. Contemporary politics revolve around the division of the spoils, and the interaction between ballot box rivals turns from competitive to collusive once legislators have been seated. While spoils motivate political life, the focal point of spoils politics has shifted from the national capital to the regions. Decentralization has increased the importance of holding local power. Sub-national politicians eagerly divide up the substantial transfer payments that now flow from the centre. This fact of political life has been especially true in ethnically diverse areas. These relatively small administrative units receive relatively high flows of transfers, and the volume has only increased with the substantial post- Suharto boundary splitting. Bringing this back to the issue of party systems and electoral competition, ethnically diverse electoral districts tend to have high rent opportunities. Voters in these electoral districts are relatively more reliant on state employment, less likely to expect delivery of public services, and more resigned to systemic corruption. Diverse electoral districts, then, are more likely to produce clientelist politics rather than programmatic politics. In the proceeding chapters I will link these sub-national rent opportunities to patterns of political organization and electoral behaviour. Indonesia’s peculiar institutional arrangement, combined with the competition for local spoils, has unleashed a political dynamic that has fragmented the party system. While the strong nationalist norms written into electoral institutions have prevented direct ethnic and regional partisan mobilization, the indirect effect of ethnicity continues to stymie efforts to force the consolidation of the party system. 87 Chapter 4 - Candidate Entry in Indonesia In 2009, 95% of Indonesia’s 11,269 national legislative candidates failed to win a seat. Most of these aspiring politicians never had a realistic chance of getting elected. In some cases these long-shot candidacies could be written off as miscalculation. Perhaps the 7,018 aspiring politicians that lined up behind one of the country’s untested parties simply over-estimated their upstart party’s appeal. It is considerably harder to explain the 1,130 candidates who ran for parties which received less than 1% of the vote in the previous 2004 election. These parties had an established record of poor performance on the national stage. It was a safe bet that none of them would even meet the newly enacted electoral threshold that would allow them to seat national legislators in 2009. Candidates for these nationally insignificant parties were aware of the electoral risks. Despite the odds, long-shot candidates stayed in the race, typically at great personal expense. Which raises the question: why engage in a hopeless political campaign? Why spend resources where there is only a miniscule chance of success? We cannot write these long-shot candidacies off as simply a lack of political experience or the unintended by-product of the nation’s under-funded education system. If this was the case the tendency to engage in hopeless political campaigns would be uniform across the archipelago. Yet some districts saw far more hopeless candidacies than others. By taking the number of candidates-per-seat as a measure of the number of hopeless candidacies we observe considerable variation in behaviour across electoral districts in 2009, from a high of 31.7 candidates-per-seat to a low of 15.4. Examining this cross-district variation can help us uncover the motivation behind candidate behaviour. Why do candidate entry rates vary across electoral districts? I argue variation in the number of candidates is (partially) explained by variation in ethnic diversity. The more ethnically diverse a district, the more candidates it will have. Existing comparative literature suggests multiple causal pathways connecting ethnic diversity and candidate entry, but none them explain why hopeless candidates in diverse districts would invest their efforts in a campaign. I offer a novel explanation that emphasizes the relationship between sub-national rents and national candidacy. Ethnic diversity produces rent opportunities, which increase the payoff from holding sub-national seats (more 88 patronage, corruption, etc). Political candidates benefit from (potentially) gaining office and gaining influence within a party. Sub-national rents increase the payoff a national candidate receives from building partisan influence: where local rent opportunities are high, it is financially beneficial to have influence with a locally powerful network. Thus ethnic diversity produces rents, which draws in candidates. My argument places candidate behaviour in a new light. Influence within even a minor party is valuable because these parties can still hold sub-national power. For instance, the 10 minor parties that received less than 1% of the vote in 2004 may have elected only 4 national legislators among them in that year, but they also controlled 851 municipal seats. Aspiring politicians are willing to take on hopeless national-level candidacies because it allows them to build and maintain influence with sub-national co- partisans. To understand outcomes at the national level, then, we need to examine sub- national dynamics. I test the argument using a series of datasets containing information on over 86,000 candidacies. Understanding the factors influencing candidate decision-making is an important subject in its own right, but it particularly pertinent to my broader investigation into the determinants of party system size. The number of candidates affects electoral outcomes. A high number of candidates on a party’s list indicate the party has many activists canvassing for support within a district. This raises the party’s profile and increases the chance a voter will support a party simply because they like the local candidate. Accordingly, when a district contains an abundance of candidates spread over a large number of parties, the dispersion of the vote fragments the party system. The chapter proceeds as follow. Section 2 presents the empirical puzzle of cross- district correlation between candidate entry and ethnic diversity and goes on to review the explanations for this phenomenon found in existing literature. Section 3 outlines my Rent Opportunities model of candidate entry, highlighting the differences between my argument and two alternative models – Communal Voting and Strategic Parties – that predict a similar correlation between candidate entry and ethnic diversity. Section 4 provides the necessary background on the Indonesian case. Sections 5 tests the Rent Opportunities model and explains why this approach is more persuasive than the potential alternatives. I conclude with a summation of the findings. 89 Why enter? Existing explanations Empirical puzzle I begin with a stylized fact: diverse electoral districts have more legislative candidates. Examining aggregate candidate numbers in Figure 8 and Figure 9, we observe a correlation between ethnic diversity and candidate entry in Indonesia’s 2004 and 2009 elections.114 Why does ethnic diversity correlate with a high number of candidates? The exact same parties compete across all of Indonesia’s districts and none of the parties have an explicitly ‘ethnic’ platform. This variation in the number of candidates competing in an election is important because it directly affects the democratic experiences of voters and the political fortunes of parties. Figure 8 – Candidates-per-seat in 2004 114 The relationship does not exist in 1999, the countries first post-Suharto contest. There are four reasons candidate entry was anomalous in 1999: 1) the lead up to the election was marked by transition era polarization; 2) the pre-election organizing period was attenuated due to the unexpected timing of the election call; 3) the election preceded the decentralization of authority to sub-national units; 4) the recruitment of candidates preceded legislation freeing civil servants from their commitment to support Golkar, the Suharto regime’s party of hegemonic control. 10 15 20 25 Ca nd ida tes -pe r-S ea t (A gg reg ate ) 0 .2 .4 .6 .8 1Ethnic Fractionalization (0-1) Candidates-per-Seat by Ethnic Fractionalization - 2004 90 Figure 9 – Candidates-per-seat in 2009 The relationship between candidate entry and ethnic diversity is consistent with my theoretical argument that rent opportunities attract politicians. There are, however, two alternative theoretical stories that can account for the relationship, which I term the Communal Voting model and the Strategic Parties model. Each model explains the correlation between ethnic diversity and candidate entry with a distinct causal mechanism. As such, they generate distinct observable implications that allow us to disentangle which of the three models is playing out in the Indonesian case. In order to identify mechanisms and observable implications it is first necessary to review three broad approaches used when accounting for variations in candidate entry. I then explain how these approaches are blended to produce two existing models that posit a relationship between entry and ethnic diversity. Approaches to entry The comparative literature offers three broad approaches to candidate entry. First the strategic candidates approach focuses on the decision-making of individual actors. Models typically contain similar assumptions. First, running for office is costly. Second, successful candidates receive a payoff if they win office, both because they get to enjoy the perks of office and they get to enact their preferred policy. Third, the expected payoff 15 20 25 30 35 Ca nd ida tes -pe r-S ea t (A gg reg ate ) 0 .2 .4 .6 .8 1Ethnic Fractionalization (0-1) Candidates-per-Seat by Ethnic Fractionalization - 2009 91 from office is higher if the candidate believes she will win. Aspiring politicians are more likely to become candidates when they think they can win office, when the payoff from holding office is large, or when the costs of running are low.115 Empirical work on candidate entry has not kept pace with theoretical advances however.116 A second strategic parties approach starts with the assumption that electoral competition in modern democratic countries is structured by political parties. Parties have an interest in winning seats and achieving policy goals. In order to maximize the number of seats and maintain internal discipline, parties tightly regulate access to the ballot for those wishing to use the party’s banner. Parties set candidate numbers to avoid coordination failures which occur when seats are lost due to co-partisans splitting the party’s vote. Coordination issues are of greatest concern to parties in plurality systems, where the costs of failure are higher (Cox and Rosenbluth 1994). Still, even parties operating in proportional systems must balance the different demands of activists and voters. These centralized decision-making processes affect the number of candidates that are allowed to enter the political sphere. Third, the sociological approach emphasizes the issue of candidate supply (Norris 1997). A citizen’s decision to become a candidate is structured by the social environment in which she lives. Factors in the broader social environment make entry more or less likely and variation in social structure explains variation in aggregate entry decisions. Variation in candidate supply has two potential effects. First, it has an impact on the composition of a candidate slate. In this case the dependent variable captures some aspect 115 A robust formal literature focuses closely on expected payoff of enacting preferred policies (Besley 1997; Feddersen, Sened, and Wright 1990; Morelli 2004; Osborne and Slivinski 1996). Consistent with established Downsian understandings, policy competition occurs across a uni-dimensional space where the citizen’s policy pay-off is determined by the distance between their policy preferences and that of the winner. These models can be tweaked to arrive at equilibrium predictions on the number of candidates entering the race, revealing the conditions under which Duvergerian outcomes should – and should not – be expected to occur. 116 There are some noteworthy exceptions. The American based literature on candidate entry uses a variety of factors to predict when ‘quality’ candidates enter a race and/or when incumbents decide to retire. In short, quality candidates enter when there is a strong chance of winning and are deterred when the probability of winning is low. See: (Carson 2005; Jacobson 1989; Stone and Maisel 2003). A similar stream of literature from Japan accounts for electoral fragmentation using a strategic candidate model. Reed argues that the ‘M+1’ outcome in Japan during the period of SNTV occurred because candidates choose to retire after several electoral defeats (Reed 1990, 2009). This process of learning winnowed the electoral field. As in the American case, it is individual candidate calculations about the probability of winning that motivate entry (and exit). 92 of candidate traits (e.g. gender, political experience, wealth). For example, a high rate of women’s participation rate in the workforce increases the ‘supply’ of potential female politicians and is positively correlated with the percentage of women candidates run by a party (Kunovich and Paxton 2005). Second, supply variations can affect the raw number of candidates. For instance, a Farmer’s Party should have trouble finding candidates in an urban environment where we expect the supply of farmers to be low. Thus sociological variables should interact with party-level variables to impact the composition and size of candidate lists. Ethnicity and entry Previous work can be leveraged to explain the correlation between ethnic diversity and a high number of candidates. First, the Communal Voting model blends sociological factors with the assumptions from the strategic candidates approach. The basic intuition from the strategic candidates approach holds that aspiring politicians will enter when: 1) they see an electoral niche they think they can fill; 2) the electoral niche has reasonable chance of being translated into a seat. The number of electoral niches may be structured by sociological factors, in this case the level of ethnic diversity. If we assume that voters prefer supporting co-ethnics, then diverse electorates should provide more electoral niches that can be filled. For instance, a district with an ethnically homogenous population will offer fewer potential support bases than a district equally divided by four ethnic groups. Aspiring politicians will enter until all the viable electoral niches are filled leading to a greater number of candidates in diverse districts. The Communal Voting model, then, presents a simple causal story: Ethnic diversity →  # of candidates How much of an effect ethnic diversity has on candidate entry is attenuated by additional factors such as demographic structure, the existence of non-ethnic cleavages, and political institutions. These factors will also determine whether candidates from an ethnic group coordinate on specific ethnic parties or spread themselves across a number of multi-ethnic competitors. Still, the fundamentals of the model – self-interested candidates and a divided electorate – are straight-forward and underlie much of the comparative 93 party systems literature (W. R. Clark and Golder 2006; Cox and Amorim Neto 1997; Ordeshook and Shvetsova 1994; Stoll 2007).117 In a second causal story, strategic parties have incentives to recruit more candidates in diverse areas. Assuming competition takes place in multi-member districts, ethnic divisions can encourage broadly aggregative parties to seek to mirror the multiplicity of social divisions within their candidate lists. Facing a diverse electorate, the party may earn an electoral boost from its ability to construct a diverse electoral slate of candidates. Extra candidates allow parties to increase the ethnic representativeness of their party lists and connect with distinct communities. Presenting voters with a ‘rainbow’ list of legislative candidates allows the party to appeal to voters across the spectrum: Ethnic diversity → ‘Rainbow’ Lists →  # of candidates In this Strategic Parties model it is deliberate partisan electoral strategy that leads to more candidates entering the race in diverse areas. The application is more limited than the Communal Voting model: the Strategic Parties model assumes both multi-member districts and parties that want at least the appearance of crossing ethnic divides. It is, however, a story that finds empirical support in post-Independence Indonesia, which makes it particularly relevant for my purposes.118 While both stories are plausible, there are important differences and over-sights. The treatment of candidates across the two models diverges sharply. In the Communal Voting model, candidates run when they think they can win. Aspiring politicians do not knowingly engage in a hopeless campaign. In the Strategic Parties model, candidate decision-making drops out of picture altogether. It does leave open the possibility of hopeless candidacies but does not explain why any candidates would take on the burden. Neither model does particularly well explaining candidate behaviour in Indonesia 117 The argument is only made implicitly. Explicit discussion of ethnic diversity and the number of political competitors tends to take place at the level of the political party (G. Cox and Octavio 1997, 152-3; Stoll 2007, 1443-1446). 118 Herbert Feith describes the logic of long candidate lists in his classic study of the 1955 elections: [I]n particular areas of an electoral district, and among different social, ethnic, and clan groups in it, the parties campaigned in terms of the attributes of the individual representatives of those whom they had included in their lists, usually in lower positions. The relatively easy procedure of nomination and the great length which lists were permitted to have encouraged the candidature of many persons who could not possibly be elected but whose name could be useful to the parties in their campaigning among particular groups of voters. (Feith 1957, 17-18) 94 because both stories assume the ultimate goal of the strategic actor is limited to attaining victory in the immediate election. By placing candidate behaviour in a broader context of multi-level competition we can understand why politicians would knowingly run a hopeless campaign. Rent opportunities and entry The basic intuition of what I refer to as the Rent Opportunities model is that Indonesian politicians are motivated by the possibility of accessing rents, and they are more likely to enter in areas where rents are high. These areas tend to be – but are not necessarily – ethnically diverse. The argument is based on a series of propositions. First, rent opportunities are high in ethnically diverse electoral districts. Second, high rent opportunities increase the payoff from holding sub-national office. Third, candidates receive a payoff from building partisan influence. Fourth, the pay-off from partisan influence is determined by sub- national office benefits. Taken together I expect high candidate entry both at the national and sub-national level where there are high rent opportunities. Sub-national candidate entry increases due to high payoff from office holding, while national entry increases due to high payoff from partisan influence in high rent areas. I unpack each claim below. Rents and diversity Rent opportunities are determined by two factors: 1) state resources available for manipulation; 2) constraints against rent-seeking behaviour. A combination of low constraints and extensive state resources make for high rent opportunities. Both resources and constraints are affected by sub-national ethnic structures. Internal and external processes increase particularistic spending in diverse contexts, expanding resources available to sitting politicians. In Indonesia, the correlation between rent opportunities and ethnic diversity long preceded the transition to democracy. It is my contention that legacies of sub-national rents exert an influence on candidate behaviour independent of communal political preferences. The effect of diversity on partisan elites and electoral outcomes is indirect, working through the mechanism of rent opportunities. 95 Office benefits, party influence, and entry Moving to candidate strategies, I start with Samuels’ basic entry model (Samuels 2003, 15). In his simple formalization, Samuels focuses on three important factors: the probability of winning, the value of the office, and the costs that will be incurred.119 I propose office benefits are high in areas with high rent opportunities. When constraints on the behaviour of public officials are loose, sitting legislators can earn a windfall through such illicit activities as budget skimming, influence peddling, and contract rigging. And when the state dominates the economy it is important to have some influence over the crafting of regulations and the distribution of jobs. In short, political office is more lucrative when rent opportunities are high. The high office benefits draw ambitious citizens into political careers and induce higher levels of candidate entry. Next, I suggest that many candidates have multiple goals. First, they want to win office. Second, they want to build and maintain valuable network connections. In other words, they want to build influence. It is this second motivation that explains the correlation between entry rates and ethnic diversity in Indonesia. These network connections can come in many forms: occupational, kinship, partisan, etc. The key point is that it pays to have connections with a locally powerful network in high rent areas. Running under a party label as part of a broader local team allows ambitious rent-seekers to solidify their network connections. Spending time and money in service to the party ingratiates a candidate with her co-partisans. For the national candidate, party influence is earned at multiple levels. As a national candidate, the national party office recognizes a candidate’s service to the party. Even an unsuccessful candidate has at least some hopes of calling in favours from party headquarters. The national level candidate also interacts with the sub-national branches that are contained within her district. Local co-partisans in office or party positions may recognize, and later reward, a candidate’s service. What is the payoff for building party influence? On the one hand, a candidate may be able to extract future positions within the party. A branch leadership position or a prominent spot on a future list of candidates are two potentially valuable commodities that influence can attain. A candidate may also try to extract state resources from sitting 119 Samuels’ model formal model is described in more detail in Chapter 2. 96 co-partisans. Favours could include contract for a friend’s construction firm, a government job for a family member or a road for a formerly supportive community. If co-partisans have attained office, party influence provides a chip even the unsuccessful candidate may seek to cash in. We can thus add on to Samuels’ model of candidate entry a payoff for building influence. The relationship can be expressed as: Ui (Running for Office o) = PioBio + Ii – Cio. “I” is the benefit that candidate i gains from building party influence by running for office. Influence itself can be further disaggregated into local and national components, and can be expressed as: Ii = NpiBon + LpiBol, where “N” is the legislative strength that individual i's party is expected to attain at the national level, multiplied by the benefit attained from holding national office; “L” is the legislative strength that individual i's party is expected to attain at the local level, multiplied by the benefit of holding local office. Several relationships can be drawn from this simple model that can help explain cross-party and cross-district variation in the number of candidates. First, as a party’s expected share of seats in the national legislature increases (Npi), the payoff from attaining influence in this party increases. Thus parties that are expected to do well nationally should attract more candidates. Second, as a party’s expected share of seats in the local legislature(s) increases (Lpi), the payoff from attaining influence in this party increases. As such, parties that are strong at the local level should attract more candidates. Third, as the benefit from local office increases (Bol), the payoff from attaining influence with a party increases. Consequently, in national electoral districts where the benefits of holding local office are high, more candidates should run at the national level. For the rent opportunities model, office benefits drive entry decisions. The presence of rent opportunities is the necessary link in the chain leading from ethnic diversity to increased number of candidates: Ethnic diversity → Rent Opportunities →  # of candidates The model can account for link between ethnic diversity in a situation like Indonesia where parties have a restricted ability to cater to ethnic demands. I have thus returned to 97 the prominent piece of folk wisdom ada gula, ada semut120: like ants, Indonesian candidates tend to congregate in areas where there is a sweet payoff for their efforts. Case background: selecting candidates in Indonesia Partisan context All legislative candidates in Indonesia must be nominated by a political party.121 There are a few restrictions on who can become a national candidate; namely people who fall under of the following four categories: 1) under the age of 21; 2) with a recent criminal history; 3) a past association with the communist party; 4) lower than a high-school education.122 Since 2004, most parties tend to conform to the non-binding regulation that 1/3 of all candidates in each district be women. Parties also have their own additional regulations. PPP, for example, limits candidacy to Muslims while Partai Golkar tries to select candidates with a five year history of party membership. Adherence to both official and partisan rules can be bent, however. Across all parties the process of selecting candidates tends to be centralized. National offices collect and submit candidate lists and thus always get the final say on list composition. There are no residency requirements on national candidates, so parties are free to place any candidate they like in a given district.123 Most parties, however, have mechanisms to solicit suggestions and feedback from sub-national branches, which include formal quotas for sub-national favourites (e.g. PAN) informal mechanisms of consultation (e.g. PPP), and membership surveys (e.g. PKS). Candidate selection timelines vary across parties but typically start around one year before the election.124 For the aspiring politician candidature is prompted by a 120 Roughly translated: where there is sugar, there are ants. 121 The one exception is for the DPD, which will be taken up in more detail below. 122 There are additional requirements, such as proficiency in the national language (Bahasa Indonesia) and ‘faith in God the Almighty’ that are generally less salient. One notable exception exists in Aceh, where local Quran recitation requirements weed out some of the secular-minded. Approximately 6% (81) of all applicants for provincial candidacy in Aceh were barred from participation in 2009 for their alleged ignorance of the Quran (Nov. 11, 2008, Jakarta Post). 123 Legislative candidates at the provincial and municipal level are formally required to reside in the province or municipality. 124 The description of the process refers to both the 2004 and 2009 elections. Due to time constraints, the 1999 elections involved a comparatively more chaotic process. 98 combination of self-selection and network pressures for involvement.125 Aspiring national candidates can apply directly to the national office or lobby their local branches for a recommendation. Lobbying processes involve both over-the-table gift-giving and large under-the-table donations. Costs associated with attaining a spot on a party list tend to vary by party size and list position (Syamsuddin 2005). The General Elections Commission (Komisi Pemilu Umum, KPU) requires a Temporary Candidates List (Daftar Caleg Sementara, DCS) of all national candidates approximately eight months before the election. Most parties pass their lists in with very little time to spare before the deadline. The KPU then examines the list to ensure that candidates meet regulations and are not registered in more than one electoral district. The reviewing process lasts about six weeks, and this period provides candidates with a last chance to weigh their political fortunes. Some political attrition occurs.126 At the end of the review process the parties submit their Fixed Candidates List (Daftar Caleg Tetap, DCT). This list stays largely stable between its public release and the printing of ballots. Campaigns and partisan influence Indonesian electoral campaigns are increasingly candidate-centered. Candidates pay most campaign expenses and, as a result, have considerable autonomy. This does not mean every candidate is an island; links with co-partisans exist. The type and strength of bonds with co-partisans varies. The strongest connections are close pre-existing bonds like family. A candidate with family members running at different levels is common. For instance, Rudolph Pardede, the former provincial PDI-P leader in North Sumatra, managed to have his son-in-law and daughter nominated at the national and municipal levels respectively (Jakarta Post, Dec 3, 2003). Nurdin Manurung, the controversial North Sumatra activist and leader of National People's Concern Party (Partai Peduli Rakyat Nasional, PPRN) provincial list in Medan, shared clan connections and a history of political activism with Sujono Manurung, the second-ranked PPRN for the Medan 125 The process of putting oneself forward can sometimes be more ‘network pressure’ than ‘self-selection.’ Top party officials are expected to run. Parties will sometimes court star candidates. While systematic data on gender and recruitment is lacking, field interviews suggest women candidates are also more likely to be actively recruited. 126 For some candidates the DCS provides a concrete measure for where they stand in the party. The most prominent defection of the 2009 campaign season involved a longstanding PDI-P legislator (Permadi) bolting to the up-start Partai Gerindra in part because he received an undesirable list position. 99 district at the national level. South Sulawesi’s powerful Limpo clan maintain fluid partisan loyalties while competing across several levels of government (Buehler & Johnson Tan 2007). Not all pre-existing bonds are as thick as blood, though. In 2004, PPDK’s North Maluku branch was built on a network of those connected with the Sultan of Ternate. Pioneers’ Party (Partai Pelopar, PP) North Sumatra organization had a concentration of politicians originating from the island of Nias. In these cases candidacy is, in part, a costly signal of an individual’s commitment to the broader pre-existing network. The very process of campaigning helps forge bonds between non-related co- partisans. Co-partisans invite each-other to common events. Also, co-partisans occasionally produce joint campaign material, allowing them to cut-down on the fixed costs of running a campaign. Figure 10 and Figure 11 include examples of such partisan teamwork. This practice of ‘running in harness’ tends to integrate co-partisans vertically, as posters and name cards are likely to have only one candidate from a given level of government. It also allows the candidates to market themselves as a team, placing them in a broader network than can assist a supporter once in power. 100 Figure 10 – PAN candidates running in harness in Medan Two candidates from Partai Amanat Nasional. Mulfachri Harahap (left) is running for the national legislature (DPR), Hapcin Suhairy, SE (right) is competing for a seat on the Medan municipal council (DPRDII). Photo taken in Medan February, 2009. Figure 11 – PPNUI candidates running in harness in Medan Two candidates from Partai Persatuan Nahdlatul Ummah Indonesia. Drs. H. Nurdin Nasution (left) is seeking a seat in the national legislature (DPR), H. Z. Zailnal Arifin (right) is competing for a seat in the provincial (DPRD) legislature. Photo taken in Medan February, 2009. 101 These short examples demonstrate one underlying point: candidates can work in teams across different levels of government. Candidates not only care and benefit from their individual run for office; they also care and benefit from their team’s broader success. By placing individual entry decisions in their partisan and social context we gain leverage in explaining candidate entry rates across electoral districts. Theory testing: entry across electoral districts Statistical testing is broken into three sections. First, I demonstrate the validity of the argument by establishing the crucial correlation between candidate entry and ethnic diversity. Second, I establish the plausibility of the rents-based mechanism by linking entry rates to rent opportunities in the absence of ethnic diversity. Third, I explore further observable implications of the Rent Opportunity model vis-à-vis other alternative explanations for candidate entry. Ethnic diversity and entry The first step in the empirical analysis is to rigorously establish the relationship between number of candidates within a district and ethnic diversity. The key dependent variable is the number of candidates run by each party in each electoral district. Comparing the raw number of candidates is appropriate when district magnitude is uniform; however, there is the potential that district magnitude could overwhelm all other variables when it varies. Instead of raw aggregates I construct a simple dependent variable termed candidates-per- seat by simple dividing the number of candidates by the district magnitude.127 The key independent variable is the ethnic fractionalization of electoral districts. To construct fractionalization measures I use data from the 2000 census conducted by the Central Statistics Agency (Badan Pusat Statistik, BPS). The 2000 census asked respondents to self-identify their ‘suku bangsa,’ which roughly translates as ethnicity. Census forms did not provide any a priori categorization, so ethnic identities were freely chosen by the respondent. Ethnic categories are reported by municipality and I use these municipal totals to construct 0 to 1 measures for all electoral districts. 127 All presented models are robust to the use of raw aggregates measures. 102 Other factors affect entry rates. District magnitude can affect a candidate’s perceived probability of attaining a seat and the status value they attach to their position. From a strictly mechanical view, a candidate anticipating a strong personal vote should be more willing to enter the race in a high magnitude district. Because electoral quotas are lower in high magnitude districts, a candidate would need to achieve a lower percentage of the electoral vote to secure a seat. Thus a candidate anticipating their personal vote will equal 10% of the total electoral vote should feel confident of winning a seat in a district with 10 seats (and a 10% quota), whereas they may not feel confident in an electoral district with 3 seats (and a 33% quota). In this case, an increase in district magnitude should increase candidate entry rates.128 Yet this electoral logic may not be the only calculation an aspiring candidate makes when evaluating entry decisions. Candidate list positions have status value, and the scarcity of positions may increase the status of holding such a position. Thus, it could be more desirable for a candidate to take a list position of slot 3 in a district with a magnitude of 3 than a list position of slot 10 in a district magnitude of 10, simply because being 1 of 3 is more exclusive than being 1 of 10. Furthermore, if candidates expect to benefit from the influence they build within the party, it is better to share the benefits with a smaller number of co-partisans rather than a larger number. If this logic motivates decision-making, district magnitude increases should decrease the value attached to candidacy and thus candidate-per-seat should negatively correlate with district magnitude. No matter the direction of the effect, district magnitude should impact entry decisions. Second, urbanization is important for two reasons. First, urban residents are likely more politically involved. Education levels are higher and the lack of agricultural duties increases leisure time. Also, cities tend to attract the politically ambitious. Beyond the sociological reasons, urbanization may also reduce the costs associated with candidate campaigns. In interviews, candidates reported that campaign spending tends to be higher 128 The force of these calculations is likely affected by the electoral rules. Even before the court mandated switch to an open list system in 2009, reforms had lowered the percentage of personal votes required to bypass co-partisans with a more advantageous list position. This should have made the previously unviable list positions seems more attractive, especially in high magnitude districts. 103 in rural areas.129 Rural voters are more likely to ask for gifts and, given the dispersion of the population, candidate gifts are difficult to target in a way that affect a high concentration of voters. The provision of goods to a multiplicity of modestly sized villages drives up costs. Urban voters, on the other hand, are more likely to be swayed by non-patronage based appeals.130 Additionally, urban voters are more exposed to media campaigns organized by national and branch partisan offices. Again, the prediction is that lower campaign costs should increase the probability of entry. To measure urbanization rates I rely on BPS statistics from the 2000 census as made available by Suryadinata et al. in Indonesian Electoral Behaviour (Suryadinata, Ananta, and Arifin 2004) Third, economic factors influence entry decisions. The direction of the relationship, however, is difficult to predict. On the one hand, wealthier districts have more citizens able to cover campaign costs. On the other hand, campaign costs may be lower in poor districts as voters are satisfied with cheaper gifts. In other words, poverty could reduce the pool of aspiring candidates while decreasing potential campaign costs. I used official poverty rates to measure economic conditions. In some instances poverty rates and per capita income levels can diverge substantially, especially in areas with high natural resource wealth, thus average income does not reflect actual conditions. For the 1999 and 2004 elections I rely on data from Suryadinata et al.; for the 2009 election I use data from BPS’s Poverty Data and Information 2007 (Badan Pusat Statistik 2008; Suryadinata, Ananta, and Arifin 2004) Two additional variables take into consideration a district’s religious context. Both secular vs. religious conflicts and traditionalist vs. modernist divides continue to animate political competition. Religious parties face distinct candidate supply issue. Whereas Indonesia is largely Muslim, some of the outlying districts have Christian or Hindu majorities. To take this issue into account I add a district-level variable to capture the percentage of Muslims in the district, a party-level dummy to capture whether or not 129 Interview with PPP DPR candidate, Medan, (11 March 2009); interview with PD DPRD candidate, Medan, (4 March 2009) 130 One candidate suggested that the urban voters’ relative preference for political programs may reflect differences in educational levels. Interview with PPP DPR candidate, Medan, (11 March 2009). 104 the party can be categorized as ‘Muslim,’ and a variable to capture the interaction between the percentage of Muslims and Muslim Party.131 The last set of party-level variables captures partisan electoral strength. The logic is two-fold. First, the Rent Opportunities model holds that participating in electoral efforts allows candidates to build influence within a party, and the payoff for building influence is more valuable when the party is either nationally or regionally strong. Thus candidates should be more likely to join parties that are strong either nationally and/or locally. Candidates, however, may also want to join locally strong parties for purely electoral reasons. By joining a party with a strong history of local electoral performance, a candidate increases his/her perceived probability of winning a seat. Likewise, a list position with strong national parties is perceived as more valuable as candidates could take advantage of a nationally recognized partisan brand. Thus a variable capturing perceived local and national strength should positively correlate with the number of candidates-per-list. To measure perceived strength I utilize past strength from the previous electoral contest.132 Results I use ordinary least squares regression, clustered by party. Results appear in Table 4, Model 1. For reasons of space I present only results from 2009.133 As expected, the relationship between ethnic fractionalization and number of candidates is positive and strongly significant.134 District magnitude has a significant negative relationship with candidate numbers, while urbanization induced higher entry. Party level variables are signed in the predicted direction and statistically significant. The supply of Muslims increases the number of candidates that run for a Muslim party. 135 Meanwhile, national and local electoral strength both correlate with high levels of candidate entry. 131 Data on the percentage of Muslims was drawn from the 2000 census. Parties are categorized as ‘Muslim’ if: 1) official pronouncements denote ‘Islam’ as the basis of the party; 2) partisan origins are traceable to pre-existing religious organizations. For more, see Appendix D, Section 1. 132 Measuring past electoral strength is complicated slightly by party registration rules and changes to electoral districts. For further details see Appendix D, Section 2. 133 For 1999 and 2004 see Appendix D, Section 3. 134 All results are robust to alternative specifications of the ethnic diversity variable, namely: 1) effective number of ethnic groups and; 2) size of the largest ethnic group. 135 Limiting the sample to homogenously (over 95%) Muslim districts do not disturb the overall results of the models presented below. 105 Table 4 – Candidate entry Determinants of Candidate List Size – National Legislature (OLS Regression, Clustered by Party) Dependent Variable: Party-level Candidates-per-Seat by Electoral District Model 1 – Baseline 2009 Model 2 – Baseline Model + Patronage Model 3 – Homogenous Districts Variable Estimate (std. err) Estimate (std. err) Estimate (std. err) Rent Opportunities .003*** (.0009) .004*** (.001) Ethnic Fractionalization .111*** (.020) .046** (.021) District Magnitude -.024*** (.003) -.019*** (.003) -.021*** (.005) Urbanization .122*** (.022) .160*** (.031) .055 (.051) Poverty .049 (.063) .056 (.109) -.221 (.160) Muslim Party -.252*** (.084) -.288*** (.082) % Muslim -.167*** (.034) -.176*** (.034) Muslim Party X % Muslim .426*** (.098) .462*** (.097) National Strength 3.34*** (.773) 3.384*** (.826) 3.833*** (1.119) Local Strength .699* (.404) .698* (.384) .707 (.509) Const .616*** (.066) .576*** (.070) .491*** (.074) Obs 2812 1938 722 R2 0.3685 0.3816 0.3842 *p < .10. **p < .05. *** p <.01. To demonstrate the magnitude of the effect I run simple simulations using Clarify. With all variables set to the mean, a party runs approximately 0.535 candidates per seat. I increase the ethnic fractionalization measure by one standard deviation, from 0.47 to 0.80. This is like moving from the district of East Java IV, a district with a majority (63%) of Javanese and a minority (36%) of Madurese, to more diverse Riau I, a district with a plurality of Melayu Riau (31%) and sizable minorities of Javanese (26%), Minangkabau (12%), Batak (7%) and other smaller ethnic groups. A one standard deviation increase in ethnic fractionalization results in a 0.037 increase in the predicted candidates-per-seat, bringing the value up to 0.573. While this is a difficult statistic to interpret, the size of this effect is roughly comparable to other important district level 106 variables. For example, increasing district magnitude by one standard deviation (7.2 to 9.4) decreases the predicted candidates per seat value by 0.048. To further facilitate interpretation I examine the aggregate number of candidates within the electoral district.136 Increasing ethnic fractionalization by one standard deviation produces a 1.43 increase in the predicted candidates-per-seat. For an average electoral district containing 7 legislative seats, this would mean approximately 10 additional candidates. The relationship between ethnic fractionalization and candidate numbers is also consistent across levels of governance. I test the relationship using provincial aggregates from 2004 and 2009. Ethnic fractionalization is always positively signed and statistically significant at the P<.05 level, even with the omission of the control variables. 137 In sum: ethnically diverse electoral districts attract more national-level candidates and diverse provinces attract more provincial candidate. Rent opportunities and entry To test the correlation between rent opportunities and candidate entry I next construct a measure of rents. I start with the common assumption that rent opportunities and public goods provision are inversely related. Thus the presence of rent-seeking behaviour can be revealed by examining past public policy outcomes. One public policy outcome of relevance is the provision of infrastructure services. Local governments in which tender processes are corrupted and state funds are directed toward more particularistic forms of spending tend to do a poor job of paving roads, keeping street lights on, and providing a consistent supply of electricity. According to my simple assumption poor infrastructure quality should indicate the prevalence of rents. For a measure of infrastructure quality I rely on data generated by KPPOD (Komite Pemantuan Pelaksanaan Otonmi Daerah 2008). KPPOD uses mass surveys of business leaders to measure infrastructure quality by municipality. Recent KPPOD reports sample all regencies within a selection of provinces (15 in 2007). Using these data and municipal-level population statistics I construct KPPOD infrastructure scores for 136 Dependent variable: candidates-per-seat. Independent variable: ethnic fractionalization. Control variables: district magnitude, urbanization, poverty, % Muslim, Jakarta dummy. 137 See Appendix D, Section 4. 107 national electoral districts in 2009. Though the KPPOD data does not cover all regencies, I am able to construct scores for 51 of 77 electoral districts. The initial scores range from 44.8 to 81.1, with infrastructure quality increasing as the score increases. As an interpretive convenience, I subtract 100 by these values so that the high infrastructure scores reflect poor governance and I use this as a rough measure of rent opportunities. Figure 12 plots this relationship. The rent opportunities measure correlates with ethnic fractionalization. As ethnic fractionalization increases, the rent scores increase. Ethnic fractionalization alone can explain just over half of the variation in the rent opportunities measure. The clear relationship between ethnic fractionalization and the rents measure adds confidence in the validity of this rough operationalization of local rent opportunities. Figure 12 – District level diversity and infrastructure quality Model 2 in Table 4 presents results when the rents measure is added to the statistical model. The size and significance of the control variables are similar across Models 1 and 2. This suggests that, despite losing approximately 1/3 of the sample size, the basic relationships hold steady across both the full and slightly truncated samples.138 138 An additional test was run on the truncated dataset using only the baseline model. Standard errors and correlation coefficients in all models closely mirrored those found in Model 1, Table 1. 108 Turning to the key independent variables, the rents measure is positively signed and strongly significant. Regions with legacies of poor governance attract more candidates. Increasing the rent opportunities variable one standard deviation from the mean (32 to 41) produces a 0.028 increase the number of candidates-per-seat. As a comparison, a similar change in the district magnitude variable produces a decrease of 0.032 candidates-per-seat. In contrast, a one standard deviation increase in the ethnic fractionalization measure now produces an increase of only 0.015 candidates-per-seat. These results suggest it is not ethnic diversity itself that is directly producing more candidates but rather the intervening rents variable. The close relationship between the rent opportunities proxy and ethnic fractionalization make it is a given that the rent opportunities score will correlate with a higher number of candidates. Strong evidence in favour of the Rent Opportunities model would be if the rents variable can explain variation in candidate numbers within relatively homogenous electoral districts. To test this I limit the sample to ethnically homogenous electoral districts. Homogeneity is a relative measure. Within the 2009 dataset, mean ethnic fractionalization is 0.47 with a standard deviation of 0.34. I defined homogenous districts as those with an ethnic fractionalization score of 0.13 or less. The remaining sample consists of very homogenous electoral districts. In concrete terms, the largest ethnic group consists of, on average, 97.6% of the entire district population. There is no theoretical reason why minuscule variations in ethnic diversity across these homogenous districts could produce any change in the number of candidates. Using the previous model as a base, I examine entry in the homogenous sample. Results appear in Table 4, Model 3. Given that the truncated sample is both ethnically homogenous and overwhelmingly Muslim, I drop the ethnic and religious variables. The rent opportunities variable is positive and strongly significant. A move of one standard deviation in the rents variable produces 0.02 more candidates on a party’s list, an impact similar in size to changing district magnitude by one unit (-0.024). This effect is substantial, especially when considered in aggregate. Figure 13 presents a simple plot of the aggregative relationship between rents and candidates-per-district. The trend is clear: number of candidates-per-district increases as the rents variable increases. 109 Figure 13 – Candidates-per-seat by infrastructure quality It is unlikely that the results are driven by an omitted variable that is affecting both the key independent and dependent variables. The most plausible omitted variable is mountainous regions. Rough terrain could produce low infrastructure scores. Social isolation caused by rough terrain could lower the accuracy of citizens’ political knowledge, which could, in turn, induce higher levels of entry by long-shot candidates. The rough terrain story is complicated by the fact that all of the most homogenous districts have very high population densities (between 588 and 1543 people / KM2) and contain a relatively low number of closely clustered regencies (between 2 and 6). There is no correlation between the rent opportunities variable and population density of the electoral district in homogenous districts.139 This strongly suggests the analysis is not overlooking a simple geographic story. In sum: candidate entry rates tend to correlate with ethnic diversity, but the relationship appears to be driven by the intervening effect of rent opportunities. When a rents variable is added to the model there is a considerable decline in the effect of ethnic 139 A simple analysis reports a correlation coefficient of .002, a standard error of .004, and a T-statistic of 0.58. 110 fractionalization on candidate entry rates. Even in the absence of ethnic diversity, a legacy of poor local governance attracts higher numbers of aspiring politicians. Alternative explanations Communal voting In the Communal Voting model, ethnic diversity produces higher rates of candidate entry because aspiring politicians can strategically exploit societal divisions. In the Rent Opportunities model, diversity produces higher rates of candidate entry because partisan ties link the candidate to local rents and provide an extra incentive to pursue a political career. One implication of the Rent Opportunities model, then, is that ethnic diversity should not produce higher levels of candidate entry in the absence of partisan competition. Indonesia offers an opportunity to test this argument empirically at a similar level of governance by toggling institutional rules while holding the social environment constant. Indonesia’s upper-house, the DPD, bars parties from competition. Parties are not allowed to forward candidates and candidates are not allowed to use party symbols during campaigns. Becoming a DPD candidate does not require the costly informal lobbying associated with attaining a slot on a party’s candidate list. DPD campaigns remain largely detached from the partisan competition for the DPR. Elections for the DPD took place in both 2004 and 2009. Electoral districts for the DPD follow provincial lines and each district is allocated 4 seats. The overlap in social context between DPR and DPD electoral districts is exact in over 50% of cases. For each year I count the number of candidates competing in each district. Since the district magnitude does not vary, the raw number is used as the dependent variable. The central independent variable measures ethnic fractionalization within the electoral district; in this case, the measure is ethnic fractionalization by province. Figure 14 demonstrates the lack of a relationship between DPD entry and ethnic fractionalization using district aggregates from 2009.140 In 2004, the key variable driving the number of DPD candidates was provincial population size. In 2009, the key factor 140 Model specifications and results can be found in Appendix D, Section 5. 111 was the strategic electoral context. Provinces that had a concentrated electoral vote in 2004 attracted fewer candidates in 2009, likely because aspiring politicians saw little hope in winning a seat. Ethnic fractionalization, however, never reaches standard levels of statistical significance in any direction. Figure 14 – DPD entry across provinces There is no social dynamic inherent in ethnic diversity that produces high numbers of candidates. The finding is consistent with the Rent Opportunities story: in the absence of partisan ties there is no extra incentive to enter the political competition in diverse areas. Strategic parties For the Strategic Parties Model, the mechanism connecting ethnic diversity and high numbers of candidates relies on well-informed, deliberate elites in the national office. National elites recognize the competing demands of ethnic groups across the country and, in diverse districts, respond to these demands through careful recruitment of candidates. Candidate list sizes expand as parties bring in additional candidates that can appeal to the multitude of communal interests. This logic implies parties are aware of, and responsive to, district-level communal demands for representation. This strongly suggests the relationship between diversity and list size is driven by the strategic calculations of the larger parties. Large parties have the most active local branches, the longest institutional 0 20 40 60 80 To tal Nu mb er of Ca nd ida tes 0 .2 .4 .6 .8 1Ethnic Fractionalization (0-1) DPD Candidates by Ethnic Fractionalization - 2009 112 memory, and the financial resources to conduct cross-country polling. It is the large parties, then, that are most likely to have the capacity to respond to district-level communal demands. The Rent Opportunities model does not rely on well-informed large parties. Instead, it suggests networks of aspiring politicians use party-labels opportunistically. Given that it requires less money and effort for a network of locally-oriented rent seekers to take-over a minor party, these strategic candidates are willing to avoid the larger parties. Instead, networks take over minor parties, placing supporters on candidate lists at the national and sub-national levels. Party Strategies and Ethnic Balancing One observable implication that can be empirically investigated is the presence or absence of ethnic balancing. If major parties are adding candidates to respond to local demands, they should also be self-consciously selecting candidates based on their ethnic backgrounds. This type of balancing behaviour should be evident in the process and outcome. One method of testing this hypothesis, then, is simply to ask national elites with knowledge of selection processes if they engage in ethnic balancing. Interviews with elites from the large parties were solicited between May and June 2009. The ‘Big-7’ parties from the 2004 election – those with over 5% of the electoral vote – were pursued. I was able to attain interviews with elites from six parties.141 Questions regarding ethnicity can be sensitive in Indonesia. When raising the issue of ethnic balancing, initial questions were deliberately posed to illicit discussion of concrete examples under the assumption that official policy and unofficial practice may diverge and the latter may be more amendable to open discussion.142 In no case did respondents ever report balancing as a motivation for selecting candidates. Respondents were typically prompted on the issue more than once but no evidence of balancing was found. There is a possibility that respondents were being 141 I was unable to attain an interview with PKB. At the time PKB was involved in a pitched internal battles caused by the withdrawal of Abdurrahman Wahid, its well-known founder and spiritual leader. 142 For example: "I have spent time in North Sumatra recently. There are many different social groups there (Karo, Tapanuli, Jawa, etc). When selecting candidates for a diverse region like North Sumatra, is it important in your party to provide a balanced number of candidates from different groups…" 113 careful with the information they revealed. Ethnic jostling can be a sensitive subject in multi-ethnic Indonesia and discussions with a non-Indonesian academic may be more likely to only receive stated policy. Respondents, however, did talk openly about a range of sensitive issue, including nepotism, the perceived failure of internal selection strategies, and internal factional struggles. In light of respondent openness on other sensitive issues, the non-acknowledgment of ethnic balancing is telling. Given that elites in the major parties did not report ethnic balancing it is unlikely that the relationship between ethnic diversity and higher candidate numbers is being driven by strategic parties.143 Local Networks and Minor Parties Two additional observable implications can be tested through an examination of candidate entry rates. First, if the Strategic Parties model is correct, the relationship between ethnic diversity and candidate entry should be strongest among major rather than minor parties. Second, if the Rent Opportunities model is right then candidates should exhibit strong signs of ‘pack’ behaviour. In other words, a candidate should be less likely to join a small national party label if there are no co-partisans running at lower levels. To investigate the first observable implication I re-ran the models from Table 1, disaggregating the sample by small and large parties. I define small parties as those with less than 5% of the national vote. The relationship between ethnic diversity and candidate entry is strong and statistically significant in the minor party samples.144 A similar result is apparent when rents measure is added; toggling party size reveals that rent opportunities only has a significant effect on minor parties. Some caution is required, however. The truncated size of major party samples could impact estimates. Still, findings do suggest that local networks take over minor parties in high diversity/high rents regions. In diverse districts candidates are joining minor parties in large numbers, but are they working in groups? To address this question I compare party entry rates at the 143 Electoral rules prohibiting candidate lists from expanding much past district magnitude also complicate the ‘balancing’ logic. Major parties tend to attract a high number of candidates regardless of local conditions, and they are constrained from expanding further to meet hypothesized demands for ethnic representation. 144 Results are contained in Appendix D, Section 6. 114 national and provincial level. Candidate entry decisions at the national level should be closely connected with those at the sub-national level. I add to the baseline model a variable capturing a party’s total number of provincial candidates / total number of provincial seats. I use the provincial total for all national electoral districts within a given province. As expected, provincial-level list sizes are strongly correlated with national-level candidates-per-seat.145 The effect is mediated by party size however. In the larger parties, provincial list size are insignificant, indicating candidate entry dynamics are distinct between the two levels of governance. In minor parties the dynamics across levels are tightly coupled; candidate entry at the national level tends to follow dynamics at the provincial level. Outside of the major parties, candidates travel in packs. This sub-section finds little evidence to support a Strategic Parties story. The large parties with the capacity to self-consciously respond to local demands for representation do not report any signs of ethnic balancing. Even if they want to respond to such demands, institutional constraints prevent significant expansion of party lists. It is the minor parties that receive an influx of candidates in ethnically diverse – and high rent – electoral districts. There is strong evidence that these candidates work within networks that populate minor party candidate lists at multiple levels of governance. In sum: an investigation of the Strategic Parties model provides further support for the Rent Opportunities model. Conclusion Ethnically diverse electoral districts in Indonesia attract more candidates. Exploration of data from post-Suharto elections reveals several distinct dynamics that support a rents centered causal story. First, the construction of a rents proxy demonstrates that, even in the absence of ethnic diversity, candidate entry levels are high in electoral districts with high rent opportunities. Second, the relationship between ethnic diversity and candidate entry does not exist in the absence of partisan ties connecting national and sub-national politicians. Third, the phenomenon is not driven by strategic recruitment by large parties; 145 See Appendix D, Section 7. 115 rather, evidence suggests networks of elites take over minor parties at both the national and sub-national levels. These findings provide a vital link in the causal chain connecting ethnic diversity, rent opportunities, and party system outcomes. Electorates in diverse districts face a higher number of vote seeking politicians; more specifically, these voters choose from a menu that includes a high number of locally robust minor parties. Further chapters explore the electoral consequences of these entry dynamics. 116 Chapter 5 - Patronage and Party Switching in Indonesia Partisan careers in Indonesia take a variety of different paths. There are committed party stalwarts like Golkar’s Edward Hutabarat, who ran for the national legislature – and failed to win a seat – in three consecutive elections. There are perpetual free agents like Edi Ramli Sitanggang, who ran for three different parties in three elections. There are former partisans who struck out on their own such as Roy B.B. Janis, a two-time PDI-P candidate who ran for the breakaway Partai Demokrasi Pembaruan (PDP) in 2009. There are a few prodigal sons like the late Zainuddin M.Z., who broke away from PPP in the 2004 election to run for Partai Bintang Reformasi (Reform Start Party, PBR), only to return to his old PPP home in 2009. And, lastly, there are many who simply dip their toe in the electoral waters for one election. Despite the fact that Indonesian political careers exhibit different patterns of party affiliation behaviour, scholars are more likely to remark on the comparative loyalty and predictability of Indonesian careers than they are their variety of pathways.146 Unlike many developing world democracies, Indonesia does not see rampant party switching by sitting legislators, due mainly to institutional rules allowing party leaders to ‘recall’ sitting legislators. Yet even academic accounts of the party bolts that do occur typically describe politicians moving from large parties to small parties out of frustration stemming from internal party disagreements.147 This contrasts favourably with the rank opportunism on display in descriptions of party switching in other developing world countries like the nearby Philippines. In this chapter I suggest Indonesian party switching is more common and more opportunistic than it first appears. Methodological biases within the discipline draw attention to legislative careers. Due to data constraints, current research on party switching focuses on the behaviour of sitting legislators. However, if we want to understand the determinants of party affiliation behaviour in countries that have 146 For instance, Ufen’s insightful study of the ‘Philippinisation’ of the Indonesian party system contrasts Filipino party switching with relatively stable Indonesian careers (Ufen 2006, 25, 29). 147 See for instance Johnson Tan’s or Ufen’s account of party splits (P. J. Tan 2006, 108; Ufen 2006, 20-1). Both author’s interpretation of the specific splits are accurate and informative. I mention them only as examples of a tendency where party switching is portrayed in the context of deep internal schisms rather than simple opportunism. 117 effectively prohibited defection we need to go beyond simply looking at legislatures and consider the career strategies of both sitting and aspiring politicians This chapter asks: do party affiliation strategies vary across districts? If so, what are the determinants of party affiliation strategies? I link career strategies to sub-national rent opportunities. Candidates that are embedded in locally powerful networks with access to sub-national rents invest little in stable partisan careers. Their electoral fortunes are closely tied to their ability to credibly promise particularistic goods, allowing them to avoid the costly obligations associated with large parties. As compared to candidates in low-rent districts, candidates in high rent opportunities areas are more likely to switch their party affiliation. In areas with low rent opportunities, where electoral success is more closely tied to national to party labels, candidates tend to build stable careers over time. The chapter forms a vital link in the causal chain between rent opportunities and the number of parties. My argument proposes that party system fragmentation occurs because viable candidates join minor parties, thereby increasing the number of parties attracting electoral support. While present data does not allow me to measure ex-ante viability, a high rate of party switching within a district suggests that affiliation decisions reflect opportunistic behaviour rather than deep local attachments to party programmes. Party switching, then, is one indicator that party labels hold little local value and suggests at least the possibility for electorally viable elites to have a successful career under a minor party banner. The chapter proceeds as follows. First I review the relevant literature of affiliation and develop a model linking affiliation to rent opportunities. Second, I provide background on party selection processes in Indonesia. Third, I describe a new dataset that I created to track candidate careers over time and demonstrate the validity of the dataset in a brief cross-party analysis. Fourth, I test my cross-district hypothesis and demonstrate that candidates are more likely to switch parties in high rent provinces. Finally, the concluding section links the findings back to my over-arching argument laying out the mechanisms of party system fragmentation. 118 Party affiliation: existing explanations Party switching in the literature Party switching is a topic with normative and theoretical importance. Normatively, we expect elected legislators to represent their support base. Instances of party switching, then, raise questions about the nature of the citizen-legislator linkage. Theoretically, switching provides a window into the big question of ‘why parties?’ Why do legislators and candidates organize their activities in the way that they do? Recent literature seeks to explain the determinants of party switching, with particular focus on the institutional and party-level factors that either encourage or constrain switching. Heller and Mershon find electoral institutions shape switching behaviour, and their investigation of the Italian case finds that candidates elected in single-member districts are less likely to defect during a legislative term (Heller and Mershon 2005). Mershon and Shvetsova find that the parliamentary cycle affects a legislator’s propensity to switch parties, as office-seekers rush to join governing parties early in the legislative term while vote-seeking legislators join popular parties as an election approaches (Mershon and Shvetsova 2008). Taken together, there is strong reason to believe that legislators respond to institutional incentives when weighing their affiliation options. Party level variables also affect a politician’s propensity to switch parties. First, legislators may switch for policy reasons. Programmatic disputes with co-partisans and/or leaders can cause switching, or legislators may switch simply because another party more fully reflects their policy preferences. Also, parties with inchoate ideologies tend to lose legislators, often to other inchoate parties (Heller and Mershon 2005). The content of a party’s policy, then, partially determines whether or not they can retain legislators within a given party system. Legislators also switch parties to improve payoffs from holding office. Major parties that can control office perks attract legislators. In the Philippines, legislators tend to switch to the president’s party once elected in order to increase their access to pork (Kasuya 2008). Brazilian legislators switch to align themselves with the party of their locally powerful governor (Desposato 2009). When office benefits are valued, and one party enjoys privileged access to these benefits, we can expect switching by office- seeking legislators. 119 Despite the recent advances, our present knowledge on party switching remains limited. First, existing studies consider the decision making of only elected legislators. Consequently, they do not pick up variation in career strategies in countries that have effectively banned party-defection by sitting legislators. Though we would expect some of the empirical predictions from legislator-centered studies to carry over, a focus on candidates shifts our attention to overlooked factors such as candidate selection processes. Second, most of the theorizing so far has considered national-level partisan and institutional factors that apply across the entire case. My study is most interested in explaining cross-district patterns of affiliation behaviour and is more attuned to sub- national sources of variation in affiliation patterns, particularly the way sub-national rents affect party affiliation strategies. Affiliation and candidate goals: local vs. national strategies To link party switching and sub-national rents I must unpack why party labels are more important in some contexts than others. It is first necessary to ask why candidates join a minor rather than a major party. There are both decision theoretic and game theoretic models that offer a stylized calculus of party affiliation decisions. Although differences exist, the building blocks are the same. Utility functions for affiliation include elements of a party’s control over resources and costs associated with switching or joining. Rather than offer a new model whole cloth, I will borrow heavily from Hicken’s model of party affiliation, as this is the one most directly designed for candidates rather than sitting legislators (Hicken 2009). In Hicken’s model, candidates calculate the expected utility of coordinating in either a major or minor party. A candidate’s payoff is determined by ‘size of the prize’ for being in the largest national party and the probability a candidate’s party will be the largest. These are offset by potential costs incurred for coordination. It is assumed that the costs increase with the size of the party, such that candidates affiliating with major parties incur higher coordination costs.148 The larger the expected payoff for participation in a major party - minus the costs of affiliation - the greater the incentive to join a major rather than minor partisan option. 148 Costs can include the “real resources” involved in coordination and the “opportunity costs” of coordination such as the loss of organizational and policy autonomy (Hicken 2009, 31) . 120 One factor in Hicken’s national-level model that affects the ‘size of the prize’ is the dispersion of vertical authority. Where authority is concentrated in a strong central government, voters and elites privilege national issues and, consequently, major party labels. While the candidate may wish to avoid costs associated with major party affiliation, the benefits from joining a major party outweigh this concern. However, where the decentralization of authority causes voters and elites to privilege local issues, the costs of joining a major national party can be avoided. Decentralization of authority reduces the size of the national prize and thereby reduces the expected utility of coordinating in a large national party relative to the benefits of coordinating in a small local party. Party affiliation and rent opportunities Expanding on Hicken’s model, I argue rent opportunities affect the ‘size of the national prize.’ Though the relative centralization or decentralization of authority tends not to vary across electoral districts in a single country, and thus can tell us little about intra-district dynamics, local rent opportunities can vary considerably. Where the local government plays a dominant role in the economic lives of the citizens, and personal relationships with local politicians provide access to state resources, elite and voter attention tends to shift from national to local issues. The greater the opportunities for local rents there are, the less the expected payoff to a candidate for coordinating her efforts in a large national party. We would expect viable candidates to be more likely to join smaller parties when the size of the local prize is large. When attaining sub-national influence is the goal, candidates are more willing to avoid the costs associated with joining major parties and instead campaign under a minor party label. The localization of partisan efforts does not preclude the possibility of strong partisan organization and identification however. Whether the competition for local rents results in either affiliation with a dominant machine or opportunistic affiliation with minor labels depends upon expectations of local rent sharing. If elites and voters expect the largest local party to monopolize control over local rents, investment in the largest local party offers a significant payoff for the participant. On the other hand, if voters and elites expect local rents to be shared, the payoff for being part of the largest party is 121 significantly reduced. Candidates are compelled to join neither a large national party nor a local machine. Local rent opportunities and career strategies: a party switching hypothesis We can put all the theoretical pieces together to arrive at predictions about cross-district switching. Sub-national rent opportunities increase the relative size of the local prize. Given their local goals, candidates and voters are not automatically attracted to major party labels. Aspiring politicians feel free to ‘shop around’ and use party labels opportunistically. If the costs associated with affiliation in a particular party become burdensome, candidates change affiliation. This logic leads to three inter-locking hypotheses.149 The first pertains to the overall propensity to switch: H1. High rent opportunities are positively associated with the number of candidates switching parties between elections. The same logic can also be applied to career building. According to the theoretical argument, there is little incentive to build a career in a party when voters disregard party labels. This leads to Hypothesis 2: H2. High rent opportunities are negatively associated with the number of candidates running for the same party in consecutive elections. Combining both arguments leads to Hypothesis 3: H2. High rent opportunities are positively associated with the proportion of candidates that switch parties relative to those that run in the same party in consecutive elections. These hypotheses linking clientelist appeals and party switching are consistent with previous findings from the comparative literature. For example, Mershon and Heller find that legislators from Southern Italy are more likely to switch parties. They suggest this phenomenon occurs due to the long standing pattern of southern politicians cultivating an apartisan clientele (Heller and Mershon 2005, 549). Despasato finds that politicians in districts where they are able to build a personalistic appeal are more likely to switch (Desposato 2009, 76). There is a small body of findings that already support the hypothesis that clientelistic politicians tend to use party labels opportunistically. In the 149 The hypotheses all assume expectations of local rent sharing. 122 empirical section I will test this hypothesis in Indonesia through an analysis of candidate careers. Before doing so, I will lay out the necessary background for the Indonesian case. Case background: selection processes in Indonesia Using original fieldwork, this section describes affiliation processes from two vantage points: 1) the national office view; 2) the candidate’s view. I pay particular attention to issues pertaining directly to my causal story, including the relative power of prominent local notables and the variety of career trajectories available to aspiring politicians. Also, a description of party-level selection processes lays the groundwork for later large-N cross-party analyses of affiliation behaviour. Selection processes: the party’s view Discussion of party affiliation strategies requires some background regarding the national parties. In all parties, selection processes for national candidates are centralized. Centralization is in part a product of the electoral system. Law requires parties to produce a single list of all national candidates for submission to the KPU. Parties are expected to vet their candidates to ensure they meet requirements pertaining to medical fitness, education, age, and legal background. The law privileges the central party office, which is left with the authority to determine the list that is submitted. This requirement only applies to the national list however; provincial and regency candidate lists need to be submitted to the provincial and regency branches of the KPU. Most parties come to an agreement that authority for selection decisions should follow the principle of subsidiary, with each level of the party – national, provincial, and regency – taking the lead in their respective arenas. While the national office has authority for selection, the amount of independent effort they put into the process varies across parties. Golkar, for instance, forms a specialized committee to recruit and vet initial prospects. A preliminary list of recommendations is then passed off to a larger central committee that narrows the field 123 and determines ballot orders.150 PAN follows a similar procedure, going so far as to put certain politicians through a candidate training course to determine eligibility.151 Decision-making authority tends to lie with a larger selection committee. This includes high-ranking members of the party’s executive, often sitting members of the national legislature. The role of the party leader varies and is partially dependent on the extent to which the party relies on the personality of the leader to attract supporters. PDI- P, which is dominated by Megawati and her husband, is notorious for the amount of control the leadership exerts on internal procedures like candidate selection. In PD President Yudhoyono was not formally involved in selection but did make selected interventions, typically at the behest of a known supporter.152 Under Amien Rais, PAN’s decision-making was dominated by the party leader and a small circle of advisers.153 The formal domination of the national executive is tempered by mechanisms designed to elicit candidate selection advice from party activists. In most cases this simply involves taking suggestion from provincial branches. Parties like PAN and Golkar attempt to formalize the amount of input given to branches by setting internal quotas for branch recommendations.154 Even without quotas, party executives often must rely on feedback from branch activists. In only one major party – PKS – is there are serious attempt to include input from the wider membership. PKS members are able to fill out a survey in which they can recommend candidates and weigh in on options and voice opinions.155 Though we do not know how heavily these recommendations weigh on committee decisions, the mechanism does provide regular members with some influence. In addition to cross-party variation in selection processes, parties provide additional rules to determine eligibility for candidacy. In some cases these requirements 150 Interview with Golkar national office official, Jakarta, (5 June 2009) 151 Interview with PAN national office official, Jakarta, (10 June 2009) 152Interview with PD national office activist, Jakarta, (18 June 2009). It was reported to me that President Yudhoyono also intervened to affect decisions made regarding his son’s (Edhie Baskoro Yudhoyono) candidacy. Yudhoyono asked the committee to demote his son from a #1 ballot position to a less desirable #3 position. Apparently, the President wanted to avoid the image of nepotism. The demotion seemed to have little impact on his son’s fortunes however; in 2009 Edhie had the highest personal vote in the entire country. 153 Interview with PAN national office official, Jakarta, (10 June 2009) 154 Interview with Golkar national office official, Jakarta, (Date); Interview with PAN national office official, Jakarta, (Date). 155 Interview with PKS national office official, Jakarta, (8 June 2009) 124 reflect the social bases of the party. The Islamist PPP, for instance, limits candidacy to Muslims.156 However, the more important cross-party variation in candidate eligibility involves limitations based on party membership. These limitations often reflect the party’s approach to kaderasasi, or “cadreization.” Kaderasasi refers to the process through which members are socialized into the party organization and prepared to take on tasks for the party’s benefit. In some parties kadresasi is a process that involves education and training; in others it is does not exist or is not taken as seriously. There is an eligibility continuum, in which parties active in kadresasi (PKS, Golkar) have stricter internal rules regulating candidacy than those with weaker or unformed kadresasi systems (PD, PKB). Golkar candidates, for instance, are expected to have five years experience as Golkar members. PKS uses a tiered system in which active, established members are privileged in selection decisions. At the other end of the spectrum, parties with weak or non-existent kadresasi processes have few requirements on who can participate and do not explicitly privilege party activism. These differences come to the fore when national office activists are asked what they look for when selecting candidates. Responses from Golkar emphasize experience, both within the party and at lower levels of government.157 PKS also emphasizes a history of activism within the party.158 PD, on the other hand, places emphasis on finding candidates whom are local notables (Tokoh Masyarakat).159 There is less pressure to please an activist base, partially because it does not rely on such a base and has not bothered created one. The other major parties fall somewhere in between. In PAN, activists have a privileged position due simply to the quota system in which branches recommend candidates. The national office, however, focuses heavily on finding candidates that can appeal widely to the electorate. This includes both local notables and nationally recognized figures like popular entertainers.160 PDI-P considers a mix of vote- 156 Interview with PPP national office official, Jakarta, (11 June 2009) 157 Interview with Golkar national office official, Jakarta, (5 June 2009) 158 Interview with PKS national office official, Jakarta, ( 8 June 2009) 159 Interview with PD national office activist, Jakarta, (18 June 2009) 160 PAN took the recruitment of artists to an extreme, and was dully dubbed by the press ‘Partai Artis Nasional’ (National Artists Party). 125 attaining abilities and commitment to the party, either as an activist or a proclaimed adherence to the visi misi (Vision and Mission) or the party itself.161 Weighing the needs of the activist base against those of the need to attain votes is a difficult process for the national office. While the national office requires help from branches vetting activists and little known community figures, selection committees will sometimes have a clear idea of who the local notables are and their capacity to pick up votes. A PDI-P official explained that one of the more difficult issues that arise involves the selection of popular local candidates that lack connections to the local party activists.162 The problem of offending the base is most acute in parties with a base to offend. The national office must make a calculated risk that the backlash they will receive will justify adding an electorally attractive local notable. In the instances of local notable preference that I was told of, the national office expected voters to respond most readily to non-programmatic ties like kinship. The strategic preference for notables was apparent even in PKS, the party with the most rigorous kadresasi program. In one story from that party, a prominent local notable with few ties to the party but a large local kinship network was able to leverage his vote-seeking ability to attain a more prominent position on the party list.163 The notable’s leverage came from the credibility of the implied threat to simply take his votes elsewhere. The point is this: national office officials know there are certain districts in which local notables can carry their votes from party to party. They try to respond accordingly by sacrificing the interests of the activists, though it is obvious that that juggling these demands is a difficult task. Selection processes: the candidate’s view Aspiring politicians have complex motives for choosing their parties. I will briefly touch on four: fit, connections, cost, and viability. First, candidates want to run under a label they feel comfortable with. Indonesian campaigns do not emphasize programmatic issues, but there needs to be at least some ideological and/or sociological congruence between the party and the candidate. For example, a local candidate in Karo who happened to be a practicing Catholic told me he liked his party, Partai Kasih Demokrasi Indonesia 161 Interview with PDI-P national office official, Jakarta, (15 June 2009) 162 Interview with PDI-P national office official, Jakarta, (15 June 2009) 163 Interview with PKS national office official, Jakarta, (8 June 2009) 126 (PKDI), because it was the only one whose logo contained the image of a Rosary.164 A national candidate in Medan said he preferred PPP because it was the only party that was exclusively for Muslims.165 In other cases the candidate likes the party leader. A Gerindra candidate in Karo explained that his attraction to the party was based on his belief that Prabowo would make the best president.166 A similar sentiment could be heard from PD candidates and activists, who flocked to support Yudhoyono.167 Preferences for a leader or a policy position are the more specific things mentioned about a party, though many others are not so discriminating. When asked how candidates attained positions, pre-established connections to party activists were a common theme. Whether they were connections through a religious community168, business associates169, NGO network170, or kinship group171, potential politicians often become candidates for a party when they are asked to by someone they know. Familial connections are common. For instance, I interviewed the campaign coordinator (and brother) of an incumbent in Karo who inherited her prominent position from her father.172 One local Partai Mederka candidate I talked to had been recruited by family members in two consecutive elections under two different party labels.173 These stories are most common at the sub-national office but the pattern does carry-over to the national level. Networks frequently populate multiple list positions at multiple levels. At all levels, the initial invitation from a known person is often the start of the candidacy process. An invitation to consider candidacy in a suitable party label is often the first step, but there are additional issues of cost and viability that influence affiliation decisions. Methodologically these issues are more difficult to observe because they most directly impact non-candidates. The financial cost of becoming a candidate can be large and does 164 Interview with KSDI DPRD II candidate, Kabanjahe, (28 February 2009) 165 Interview with PPP DPR candidate, Medan, (11 March 2009) 166 Interview with Gerindra DPRD II candidate, Kabanjahe, (27 February 2009) 167 Interview with PD DPRD candidate, Medan, (4 March 2009); Interview with PD DPRD candidate, Kabanjahe, (28 February 2009) 168 Interview with Hanura DPRD II candidate, Bengkel, (6 April 2009) 169 Interview with Hanura DPRD I candidate, Pematang Siantar, (7 March 2009) 170 Interview with Barisan Nasional DPRD I candidate, Pematang Siantar, (7 March 2009) 171 Interview with Partai Mederka DPRD I candidate, Pematang Siantar, (7 March 2009) 172 Interview with PKPI DPRD II candidate’s Tim Sukses head, Kabanjahe, (28 February 2009) 173 Interview with Mederka DPRD II candidate, Pematang Siantar, (7 March 2009) 127 vary across parties. Additionally, there are potential non-economic costs. I interviewed one young man who had been a local candidate for PDI-P in Tapanulli Tengah but was forced to withdraw his candidacy.174 His position on the ballot was desirable enough to attract the attention of a wealthy aspiring politician. Key members of the local party were allegedly paid to threaten the young candidate, who withdrew in the face of harassment. The anecdote illuminates how party politics can be a costly venture, both in terms of financial resources and personal well-being. The anecdote also speaks to the issue of party viability. The young candidate was targeted because his position was desirable: there were local expectations that the PDI-P banner could attract supporters. Candidates like to join parties that are popular, or at least not hopeless. But minor parties are not always hopeless. Candidates possess some ability to sort out the difference between local and national trends. I ended candidate interviews by asking for an electoral forecast and was often told that the candidates’ party would have a weaker national performance than its local performance. This was particularly true for minor party candidates. Minor party candidates were likely to believe their electoral fortunes would stem from their own efforts. In some cases this is false hope, but where many minor party candidates exist we can surmise that locals see an open electoral market. Like motivations for party affiliation, candidate career paths also vary. Many candidates have prior experience running for office. In most cases, candidates run for the same party in repeated elections. This is particularly true for members of large parties and incumbents. In Karo, both the young Partai Keadilan dan Persatuan Indonesia (PKPI) incumbent and her father had run under the PKP banner.175 In Medan I spoke with a provincial PD candidate who had previously run as a national candidate.176 When initial investments in party labels are followed by some measure of success, either personal or partisan, candidates tend to stick around. We also find party switchers though. One candidate I talked to in Pematang Siantar reportedly switched from Golkar to Hanura because internal corruption drove the 174 Interview with former PDI-P DPRD II candidate, Medan (25 February 2009) 175 Interview with PKPI DPRDII candidate’s Tim Sukses head, Kabanjahe, (28 February 2009) 176 Interview with PD DPRD candidate, Medan, (4 March 2009) 128 cost of candidacy too high in the former.177 There was also a natural ideological fit, as Hanura itself was founded by a former Golkar presidential candidate. Other career paths are more chaotic. Discussions with a Barisan Nasional (Barnas) candidate in Pematang Siantar revealed a bewildering political history that included prior stints in PKB, PAN, and Partai Demokrasi Kasih Bangsa (PDKB).178 Indeed, prior to taking a position with Barisan Nasional in the fall of 2009, the candidate had organized and been part of the 2009 list for Partai Perjuangan Indonesia Baru (PIB). Labels for such candidates are unimportant and simply provide access to the ballot. The party switcher story related in Pematang Siantar was similar to the issues raised by national party officials when discussing the delicate process of accommodating local notables. These candidates trust their own appeal and feel free to party switch in order to achieve the best short-term deal. While the promiscuous switcher from Pematang Siantar was something of an outlier, there were similar stories occurring at the national level in the same province. PD’s headline national candidate in Medan, Abdul Wahab Dalimunthe, had previously been a legislator for Golkar. Gerindra’s Martin Hutabarat, who also managed to win a seat, had been a Golkar candidate in past election. Idialism Dachi, the incumbent for Partai Pelopor, switched to PKB before the election. Edi Ramli Sitanggang moved to PD from Partai Pancasila Patriot, though before that he too had been a candidate Golkar. And these were simply a few of North Sumatra’s big name switches. To sum up: the overview of party selection processes demonstrates that there are national-level differences in selection rules, especially as they pertain to candidate eligibility. Yet even the parties with an active membership base must deal with the challenge of incorporating and appeasing local notables who bring an independent voting block. An analysis of candidates reveals that affiliation decisions are driven by a number of factors, including cost and perceptions of local viability. Candidates follow a variety of career trajectories, with some investing in one label over multiple elections and others switching at their convenience. 177 Interview with Hanura DPRD candidate, Pematang Siantar, (7 March 2009) 178 Interview with Barisan Nasional DPRD candidate, Pematang Siantar, (7 March 2009) 129 A party affiliation dataset Generating the dataset My argument claims candidates should be more likely to switch parties in areas with high rent opportunities. Where candidates can credibly build a clientelist appeal they should be less willing to invest in party careers and more willing to hop across party labels opportunistically. Testing the hypotheses, then, requires examining candidate career paths over time. The first task involved gathering candidate lists. For 2004 and 2009 I was able to attain KPU records of candidate lists. Candidate lists for 1999 came from the Indonesian daily newspaper Kompas and were collected from archives at the Indonesian National Library in Jakarta. Given the process of manual input and the state of the record there was some minor data loss. Minor problems aside, I collected accurate records of candidate lists for three years, allowing for study of two periods: 1999-2004 and 2004- 2009. Given the substantial number of candidates under analysis I chose to use a matching program to track careers. I opted for Rikhil R. Bhavnani’s (2009) “RB- AMIN.exe: A Tool to “Fuzzy” Match Indian Names” program. RB-AMIN was designed to match databases of names in developing countries where naming practices are flexible in both spelling and order. The program analyzes each ‘source’ name against a ‘target’ database. Match scores are generated for each comparison and the program identifies all matches reaching a threshold match score. The threshold score thus separates the probable matches from the unlikely matches. To build a ‘source’ and ‘target’ database I first removed all titles from a candidate’s name. Conventions regarding the display of titles are even more unstable then naming conventions, thus removing titles increases the probability that a ‘match’ will be detected. Candidate lists were sorted by province and matched against each other across time. Matches were then examined individually to determine the plausibility of the match. I developed a decision-tree to guide the process of sorting the false matches from the probable matches.179 179 See Appendix E, Section 1. 130 For the purposes of convenience, I refer to all confirmed matches as Run-Agains. Run-Agains were further delineated into those that ran again for the same party - whom I refer to as Loyalists - and those that changed affiliation - whom I refer to as Switchers. There were numerous problems with this method worth mentioning. First, my method required more matching information to be classified a Switcher than it did a Loyalist.180 This likely produces an over-count of Loyalists and undercount of Switchers. Second, there is likely some systematic cross-regional bias in the matching process. Naming conventions in Indonesia’s less diverse Javanese areas are simpler than the more diverse non-Javanese areas. Non-Javanese often have multiple, complex names, while it is common to find Javanese candidates with only one name. As a result, there was likely an undercount of matches in non-Javanese areas and an over-count in Javanese areas. Third, the process does not take into account cases where Loyalty or Switching cross provincial lines. The province-by-province examination was necessitated by hardware and software limitations. Matching across provinces database was beyond both the ability of the matching program and the computational power I had access to. As a result, I cannot capture careers that span provincial boundaries. This produces a systematic undercount of those types of candidate likely to cross borders, namely incumbents and established activists. Fourth, my final counts were aggregated by province and thus capture no information on rates of intra-province switching between districts. This omission reflects technical limitations: aggregating datasets by province reduced the computational requirements of the project. There were interesting variations in how parties treat intra- provincial districts that can be probed in future work. For now, however, I simply assume career paths developed on a provincial basis.181 Despite these limitations, the method of career tracking still produced a dataset that could be used to test my hypothesis regarding party switching. Some of the 180 Near matches were automatically sorted as ‘Loyalists’ if they were from the same party but required additional verification to be sorted as ‘Switcher’ if they were not from the same party. 181 This assumption has some empirical validity. Parties are not organized by electoral district; rather, the provincial party branch is responsible for recommending candidates to the national office for all national districts within provincial boundaries. 131 systematic biases – such as the regional difference in naming conventions – have an effect on cross-provincial comparisons. The biases have affected the total number of matches uncovered across provinces (H1 and H2), but should not bias the relative proportion of candidates who switch parties when looking only at the total number of matches (H3). Describing the data I identified a total of 1827 matches in both time periods. Of these, 806 were matched in the first time period (1999-2004) and 1021 in the second (2004-2009). The proportion of Run-Agains increased between the two time periods. Approximately 5.9% of all candidates from 1999 ran again in 2004. The proportion of Run-Agains from 2004 to 2009 was 13.2%. The higher percentage represents two trends. First, over 6000 more candidates ran in the first election than the second. Many of the candidates in the first period were swept up by the enthusiasm of the transition. Ballot access was relatively easy to attain as the electoral authorities erred on the side of inclusion when registering parties. In 2004, fewer parties were granted access to the ballot and the enthusiasm of 1999 gave way to anti-party attitudes. Declining numbers of candidates was accompanied by a trend toward professionalization. Those that entered in the second election were more committed to pursuing a political career rather than passively participating in the spectacle of elections. 132 Table 5 – Switching and loyalty: descriptive statistics Descriptive Statistics Time Period Candidates (T1) Candidates (T2) Run- Agains (% T1) Loyalists (% T1) Incumbent Loyalists (% T1) Switchers (% T1) 1999-2004 13733 7721 806(5.9%) 609 (4.4%) 186 (1.4%) 197 (1.4%) 2004-2009 7721 11275 1021(13.2%) 801 (10.4%) 237 (3.1%) 220 (2.8%) Total 21454 18996 1827(8.5%) 1410 (6.6%) 423 (2.0%) 417 (1.9%) Most Run-Agains re-offer for the same party. The percentage of Loyalists varies slightly across periods: 75.6% in the first and 78.4% in the second. Incumbents make up a substantial portion of Loyalists. In the first period, about 30.5% of all Loyalists were incumbents. That number held steady at 29.6% in the second period. Overall, about 70% of Loyalists were not incumbents. This indicates that many candidates build their careers within a party over time and that Loyalty is not simply a function of legislative caucus size. The alternative career path involves switching parties. Between the first and second periods, the number of Switchers fell from 24.4% of all Run-Agains to 21.6%. This makes the raw number of identified Switchers relatively small: 197 in the first period and 220 in the second. Despite the small number, the proportion of Switchers can vary considerably across provinces, from a high of 50% to a low of 0%. Validity testing The methods used in the chapter are new and initial scepticism is in order. Before moving on to a cross-district analysis, it was important to ensure the validity of the data generation process. One way of testing this is to examine whether or not the party-level data fit common knowledge of the Indonesian party system. There are two ways to do this. The first is to consider Loyalty patterns across parties. Those parties with relatively more intensive kadresasi programs should exhibit higher rates of loyalty than those with very loose membership commitments. The logic is straight-forward: those candidates willing to invest time in intensive kadresasi programs should be those who build long- term party careers. Empirically, PKS and Golkar should be expected to have higher relative rates of loyalty than PKB or PD. 133 We can also test the validity of the data generation process by looking at patterns of party switching. There are several examples of prominent party splits, in which a number of prominent activists leave a major party and form a splinter party. These party splits should show up when we look at the dyadic patterns of party switching. For example, finding that a high number of PDI-P members switched to PDP between 2004 and 2009 would suggest the data generation process was accurate. If, however, the data indicated a high number of PDI-P members switched in to PKS there would signal a problem with the dataset. Validity test 1: loyalty across parties Table 6 displays aggregate candidate Loyalty data across the major parties. Separate data are presented for each time period: 1999-2004 and 2004-2009. The results conform closely to prior expectations. The two parties with the most intensive kadresasi programs – PKS and Golkar – both have the highest rate of candidate retention. The PKS result is particularly noteworthy. PKS won very few seats in 1999, which meant they had very few incumbents. In general, incumbents were more likely to be Loyalists. That PKS maintained such a high rate of candidate retention despite its modest 1999 result speaks volumes to the commitment of their cadre. Table 6 – Career strategies across time in major parties Career Strategies of Major Party Candidates: Loyalty Across Time and Parties Party Name Candidates (T1) Candidates (T2) Seats (T1) Total Loyalists (% T1) Incumbents (% T1) Non Incumbents (% T1) Time period: 1999-2004 Golkar 919 653 120 187 (20.3%) 66 (7.2%) 121 (13.2%) PKS 273 441 7 47 (17.2%) 4 (1.5%) 43 (15.8%) PDIP 643 565 153 100 (15.6%) 63 (9.8%) 37 (5.8%) PAN 551 526 34 53 (9.6%) 15 (2.7%) 38 (6.9%) PPP 690 495 58 65 (9.4%) 15 (2.2%) 50 (7.2%) PKB 616 464 51 45 (7.3%) 17 (2.8%) 28 (4.5%) Time period: 2004-2009 Golkar 653 640 128 172 (26.3%) 59 (9.0%) 113 (17.3%) PKS 441 579 45 103 (23.4%) 30 (6.8%) 73 (16.6%) PDIP 565 629 109 110 (19.5%) 53 (9.4%) 57 (10.1%) PPP 495 473 58 82 (16.6%) 19 (3.8%) 63 (12.7%) PAN 526 593 52 73 (13.9%) 25 (4.8%) 48 (9.1%) PD 435 668 57 58 (13.3%) 27 (6.2%) 31 (7.1%) PKB 464 393 52 47 (10.1%) 15 (3.2%) 32 (6.9%) 134 PKB, on the other hand, had the lowest rate of Loyalty in both time periods. This reflects the scant attention paid to party building by PKB’s leaders. In the first period PKB was dominated by the personality of Abdurrahman Wahid and relied heavily on NU, a prominent religious organization, to mobilize voters. The second period saw a nasty party split. The factional feuds, combined with the close connections to NU, effectively stunted PKB’s development as an autonomous organization. The PD story is also interesting. It had the second lowest rate of loyalty in the second period. What is striking about PD’s loyalty rate is that it is dominated by incumbents; very few non-winners ran again under the PD label. This reflected both weak kadresasi processes and the increased value of a PD slot in the second period. In 2004 PD was an upstart with a strong presidential candidate; by 2009 it was the governing party. Those candidates not elected in the first caucus were left by the wayside as prominent office-seekers flocked to the party of SBY. Perhaps even more remarkable than the loyalty patterns themselves was the stability of the patterns over time. Figure 15 plots Loyalty rates of the six major parties that participated in both time periods. The X-Axis is Loyalty rates from 1999-2004, the Y-Axis Loyalty rates from 2004-2009. A relationship is evident. In both periods the order from highest to lowest Loyalty goes Golkar, PKS, PDI-P, with PKB at the bottom. PAN and PPP switch between 4 and 5, but the differences between the two parties are minimal. Variations in Loyalty are not simply transitory; they reflect long-term structural differences across parties. The stability of this finding and its conformity to prior knowledge lends credence to the data generation process. 135 Figure 15 - Loyalty across time in major parties Validity test 2: switching and party splits The major parties have been wrecked by a number of splits. Between 1999 and 2004, PDI-P dissidents formed Partai Nasional Banteng Kemerdekaan (PNBK). Suharto’s children and other Golkar figures were instrumental in forming the Partai Karya Peduli Bangsa (PKPB). Also, many prominent PPP figures joined the upstart PBR. There were even more splits between 2004 and 2009. Disaffected members of PAN formed Partai Matahari Bangsa (PMB). Key PDI-P legislators started PDP. Golkar’s former presidential candidate started Hanura. And Partai Kebangkitan Nasional Ulama (PKNU) spun out of the chaos of PKB. A sensitive measure should pick up these party fissures. Table 7 reports the top 5 dyadic switches for each period. What is immediately noticeable is the variety in switching dynamics. Switching, especially in the first period, took place in ones and twos; there were few signs of mass movements across parties. Despite this, the matching process in first period still picked up two of the major fissures: PDI-P to PNBK and PPP to PBR. Splitting was more evident in the second period: among the top-5 dyadic switches, 3 (PDI-P/PDP, PAN/PMB, PKB/PKNU) was prominent instances of party splits. The method undoubtedly misses some of the action. It is likely that many candidates who participated in a split do so to improve their position, thus we should see many sub-national candidates become national candidates. Yet in Golkar PAN PDIP PK PKB PPP 10 15 20 25 30 Pc t L oy ali sts fro m T1 To tal (2 00 4-2 00 9) 5 10 15 20Pct Loyalists from T1 Total (1999-2004) Percentage of Loyalist Candidates in Major Parties 136 spite of the bluntness of the matching tool, we still see results that conform to our established knowledge of the Indonesian system. Table 7 – Switching by party dyad Top-5 Party Dyads for Switchers Time Period: 1999-2004 Rank Source Party Target Party # Matches 1 PKU PPP 5 2 PDI-P PNBK 4 2 PDR PPP 4 2 PPP PBR 4 5 PBB PPP 3 Time Period: 2004-2009 1 PKB PKNU 9 1 PDIP PDP 9 3 PBR Hanura 6 3 PAN PMB 6 5 PD Barnas 4 Theory testing: career strategies across districts Hypothesis 1: number of switchers across districts I first examine whether provinces with high rent opportunities produce a higher number of party switchers. To measure rent opportunities I use the proportion of civil servants in the modern (non-agricultural) economy. To construct a dependent variable that can be easily compared across provinces I divide the raw number of candidates switching parties by the number of seats in the province. Thus switchers-per-seat should positively correlate with civil service size. I also add a control for the total number of seats in the province, which I call provincial magnitude. Table 8 presents results. I run separate models for each time period. In the first time period (1999 to 2004) neither of the independent variables reaches standard levels of statistical significance. Civil service size is positively signed, however. This does conform to expectations. 137 Table 8 – Switching across districts Number of Switchers Across Districts Dependent Variable: Switchers / Provincial Seats Time Period: 1999-2004 Time Period: 2004-2009 Variable Estimate (std. err.) Estimate (std. err.) Civil Service Size 0.57 (0.49) 2.55*** (0.66) Provincial Magnitude 0.004 (0.002) 0.005** (0.002) Const 0.21 (0.13) -0.04 (0.11) N 26 32 R2 0.03 0.31 *p < .10. **p < .05. *** p <.01. The performance of the model improves markedly in the second time period. Both variables are positively signed and statistically significant. Civil service size strongly correlates with the relative number of switchers in the province. In concrete terms, we would expect an 8 seat province with a modestly large civil service size (20%) to have about 2 more party switchers than a similar province with a relatively small civil service (10%). While this may not appear large, the propensity to identify switchers is quite low: the mean number of switchers-per-province is 6.8 and the median is 3. Thus the addition of 2 switchers predicted from this simulated example is indicative of considerable change. It is also noteworthy that the district magnitude is positive and significant. This finding is counter-intuitive: we might expect low magnitude districts to be more personality based as voters are more likely to actually know the major candidates. Personality politics should produce less party loyalty. One possible explanation for this finding could be informational constraints faced by party branches in high magnitude provinces. The national office likes to fill the top spots on party lists with their favourite candidates. To fill less desirable spots the national office leans heavily on the party branches for advice. When there are many undesirable spots to fill, the branch organization may be overwhelmed by the task and suggest less-than-loyal candidates. In other words, it is easier for the Bengkulu branch of a party to suggest 2 or 3 solid candidates to round out a national list than it is for the Central Java branch to make 50 or 138 60 suggestions. As the complexity of the task increases, screening processes fall apart and branches rely on non-loyalists to fill spots. In sum: the relationship between switching and civil service size is positively signed in both time periods and statistically significant in the second. A higher number of candidates switch parties in high rent opportunity provinces. This finding provides support for Hypothesis 1. Hypothesis 2: number of loyalists across districts I next examine whether provinces with high rent opportunities produce a lower number of party loyalists. The main variables remain the same. In this case, loyalists-per-seat is the number of candidates re-offering within the same party. The hypothesis suggests the variable should negatively correlate with civil service size. Table 9 displays results. Testing finds no evidence to support the hypothesis. Neither the number of seats in the province nor provincial civil service size correlate with the relative number of party loyalists. Counter to expectations, civil service size is signed in the positive direction; however, it never approaches standard levels of statistical significance. The variables that affect party switching have no discernible impact on party loyalty. Available evidence offers no support for Hypothesis 2. Table 9 – Loyalty across districts Number of Loyalist Across Districts Dependent Variable: Loyalists / Provincial Seats Time Period: 1999-2004 Time Period: 2004-2009 Variable Estimate (std. err.) Estimate (std. err.) Civil Service Size 0.67 (0.97) 1.14 (1.66) Provincial Magnitude 0.005 (0.004) 0.005 (0.005) Const 1.04*** (0.25) 1.15*** (0.27) N 26 32 R2 -0.03 -0.03 *p < .10. **p < .05. *** p <.01. Hypothesis 3: relative proportion of switchers across districts To test the third component of the hypothesis I examine the cross-province propensity to either switch parties or re-offer within the same party. To construct a variable measuring 139 the relative propensity to switch I divide the number of switchers in a district by the total number of ‘Run-Again’ candidates. The key independent variables remain the same. Table 10 contains results from statistical testing. Consistent with prior tests, we find a positive relationship between civil service size and the relative proportion of switchers in both time periods. However, this relationship is only statistically significant in the second time period. Nonetheless, in the second period the relationship is strong. When compared with provinces with small civil services, provinces with large civil services have a considerably higher proportion of party switchers to party loyalists. Table 10 – Proportion of switchers across districts Proportion of Switchers Across Districts Dependent Variable: Switchers / Run-Agains Time Period: 1999-2004 Time Period: 2004-2009 Time Period: 2004-2009 Variable Estimate (std. err.) Estimate (std. err.) Civil Service Size 0.31 (0.24) 1.10*** (0.32) -4.98 (5.22) Provincial Magnitude 0.001 (0.001) 0.002* (0.001) 0.002* (0.001) Corruption (Avg TI score) -0.08 (0.11) CivilServiceXCorruption 1.14 (0.98) Const 0.14** (0.06) 0.04 (0.05) 0.46 (0.57) N 26 32 32 R2 0.0044 0.2375 0.2493 *p < .10. **p < .05. *** p <.01. I plot the relationship in Figure 16. On the X-axis is civil service size; on the Y- Axis is the proportion of party switchers out of total Run-Agains. The relationship is not perfect; there were obviously other factors at play. Yet the pattern is clear: propensity to switch parties across elections increases with civil service size. The civil service variable alone accounts for approximately 20% of variation in switching propensity across provinces. 140 Figure 16 – Switching across districts [2004-2009] Simulations display the size of the effect. The median district had 17 Run-Again candidates. If we hold the number of Run-Agains constant, we would expect the modestly low-rent province (civil service size of 10%) to have 3 switchers and 14 loyalists, and the modestly high rent province (civil service size of 20%) to have 5 switchers and 12 loyalists.182 In percentage terms, a relatively low-rent province should see around 18% of Run-Again candidates to switch parties; however, in a province with modestly high rent opportunities should have about 29% of Run-Again candidates to switch parties. The size of the civil service does positively correlate with the proportion of Run- Agains that switch parties between elections. The rent opportunities proxy is imperfect. There is some suggestion that it is the combination of high resources (civil service size) and low constraints that actually drive affiliation decisions.183 Nonetheless there is strong evidence that candidates are more likely to switch parties in areas with high rent opportunities as defined simply by civil service size. In sum: the available evidence supports Hypothesis 3. 182 The assumption of constant Run-Agains is not valid. Higher rent opportunities should have a higher number of Switchers, which increases the number of Run-Agains. The assumption is simply a convenience for the purposes of presentation. 183 See Appendix E, Section 2. 0 .1 .2 .3 .4 .5 Pc t C an did ate s S wit ch ing Pa rtie s 0 .1 .2 .3Civil Service Size Pct Candidates Switching Parties by Civil Service Size 141 Alternative explanation: switching and ethnicity The underlying argument of the dissertation is that the correlation between party system size and ethnic diversity in Indonesia is partially accounted for by the hidden variable of rent opportunities. Party switching demonstrates an opportunistic use of party labels that contribute to party system fragmentation. To support the underlying argument of the dissertation it is necessary to show that civil service size and not ethnicity diversity drives party affiliation strategies. I re-run the propensity to switch model and replace Civil Service Size with Ethnic Fractionalization. Results are displayed in Table 11. As might be expected, the relationship between ethnic fractionalization and party switching is positively signed in both years. However, the relationship does not reach statistical significance. I add civil service size to the model in the second time period. Adding the variable causes the ethnic fractionalization correlation coefficient to shrink considerably. Whereas ethnic fractionalization approaches standard levels of statistical significance when civil service size is omitted, the inclusion of the civil service variable reduces any suggestion of a correlation between ethnic fractionalization and switching. The civil service variable, on the other hand, remains positive and strongly significant. From this we can conclude that the relationship between civil service size and party switching is not simply an artefact of ethnic diversity. Table 11 – Ethnicity and switching across districts Ethnicity and Switching Across Districts Dependent Variable: Switchers / Run-Agains Time Period: 1999-2004 Time Period: 2004-2009 Time Period: 2004-2009 Variable Estimate (std. err.) Estimate (std. err.) Estimate (std. err.) Civil Service Size 1.07*** (0.35) Provincial Magnitude 0.001 (0.001) 0.001 (0.001) 0.002* (0.001) Ethnic Fractionalization 0.09 (0.09) 0.11 (0.08) 0.018 (0.08) Const 0.13 (0.08) 0.12* (0.07) 0.03 (0.07) N 26 32 32 R2 -0.0180 -0.0115 0.2118 *p < .10. **p < .05. *** p <.01. 142 Conclusion In this chapter I set out to demonstrate cross-district variation in party affiliation strategies. I argue that candidate affiliation decisions are influenced by local rent opportunities. When there is an expectation of local rent-sharing, the payoff for affiliating with either a large national or local party is reduced. Candidates believe they are able to gather support based on the delivery of clientelistic goods, or at least the promise to do so. This leads candidates to use party labels opportunistically. Thus candidates in high rent areas should be more likely to switch parties and candidates in low rent areas should be more likely to invest in building a party career. I constructed a dataset that tracks candidate careers over time and coded for two basic strategies: loyalty and switching. To check the validity of the dataset I cross- checked aggregate party results against established wisdom regarding party cadreisastion programs and party splits. The validity tests suggest the dataset does capture basic patterns within the broad Indonesian party system. I then conducted a cross-district analysis of candidate career strategies. Between 2004 and 2009, individual candidates are more likely to switch parties in provinces with high rent opportunities, as measured by civil service size. On the other hand, rent opportunities have no impact on the likelihood of an individual candidate’s propensity to build a career within the same party. Taken in the aggregate we find that candidates that in high rent provinces are more likely to switch parties than candidates in low rent provinces. In areas with high rent opportunities, elites and voters are less interested in party labels. Party and electoral politics are geared toward clientelistic distribution. Because party labels are relatively less important in these areas, candidates feel free to hop between parties at their convenience. The basic logic of this argument, if not its underlying causes, is recognized by national party office officials in Jakarta. Even the parties with the most extensive kadresasi programs must bend to accommodate candidates that can attract voters with clientelist appeals. And even then the accommodation may not be enough to prevent defection. These findings contribute to my argument regarding the causes of party system fragmentation. When viable candidates know they can rely on a personal, clientelist 143 appeal they are more likely to join small parties. Small parties are able to pick up votes namely due to their association with clientelistic local notables. The next chapter will examine how these viable small-party candidates are able to utilize clientelistic appeals and demonstrate that these appeals impact voting behaviour. 144 Chapter 6: Patronage, Ethnicity and the Personal Vote in Indonesia With each successive election the competition for legislative seats has become more ‘personal’ and less ‘partisan.’ Whereas many candidates in 1999 were idle and barely known within their district, by 2009 legislative campaigns had become a candidate centered affair. The evolution was prompted by institutional changes; namely, the gradual move to open-list competition that directly tied victory to personal performance and the staggered election timetable that led presidential hopefuls to hold back on advertising until after the legislative contest. In the legislative contest, spending by local candidates has now eclipsed the anemic efforts of the national party headquarters. The declining electoral value of party labels has been accompanied by an expansion of the vote attained by minor parties. Observers of Indonesia often overlook that the general increase of personalism across time occurred in a context of sharp variance in personalism across regions. The rate of ‘preference voting’ offers one way to measure the strength of the connection between candidates and the electorate.184 In 2004, preference voting rates varied widely across districts, with extremes ranging from 32% to over 82% of the electorate choosing to support an individual candidate on a party’s list. Why would individual candidate campaigns be so successful in attracting votes in some districts and ineffective in others? In this chapter I argue that voters respond to the promise of future gifts. When aspiring politicians can credibly commit to supplying favours after the election, voters are more likely to support an individual legislative candidate; when voters do not expect post-election favours, they tend to discount candidate appeals and focus on party-level factors. I link variation in preference voting to rent opportunities. The credible promise of post-election favours creates incentives for voters and candidates to form lasting relationships, which show up in preference voting rates. The personalization of politics associated with rent opportunities is an important piece of the broader causal story linking rents to party system size. Electoral 184 While Indonesia’s preference vote is a form of a personal vote, I use the term separately in the chapter. ‘Preference vote’ will be used whenever discussing the mechanisms through which candidates receive votes in Indonesia. ‘Personal vote’ will be used when discussing the broader theoretical literature pertaining to the specific efforts and appeal of an individual candidate. 145 fragmentation increases in high rent areas in part due to the success of minor parties. Rent opportunities cause an increase in the salience of personal connections and a corresponding decrease in the salience of party labels. Consequently, minor parties are able to gain a foothold because voters support their favoured patrons who happen to run under minor party labels. The chapter proceeds as follows. Section II examines the theoretical literature both defining and explaining the existence of a personal vote. Section III develops my theory linking preference voting to rent opportunities. Section IV provides necessary case background on political campaigns in Indonesia. Section V tests the argument empirically and rigorously examines the alternative explanations. Personal voting: existing explanations Defining the personal vote Definitions of the personal vote range from more narrow conceptions emphasizing the support a politician gains through personal efforts (Cain, Ferejohn, and Morris P. Fiorina 1987; Kitschelt 2000) to broader understanding which encompass all support attracted through either efforts or reputation (Carey and Shugart 1995; Marsh 2007). For this discussion I adopt the latter, more expansive definition. The personal vote has both a credit-claiming portion and an attributional portion. The credit-claiming portion refers to what a candidate has done or will do. This could include a reputation gained through distinctive policy positioning, case-work, or pork-barrel spending. The attribuitonal aspect refers to the electoral impact of who a candidate is. This could cover attributes like ethnicity, religion, clan, or gender. It also covers relevant personal qualities that can affect electoral success, such as oratory abilities or charisma. To tweak Kitschelt’s (2000, 852) definition, the personal vote is defined as the effect of a candidate’s attributes and actions on his or her electoral success, net of aggregate partisan trends that affect partisans as members of their parties. Marsh’s discussion on the stages and impacts of the personal vote is particularly relevant to this exploration of preference voting in Indonesia (Marsh 2007). Attention must be paid to a key question: does a preference for a candidate guide party choice? Personal voting can occur at two different stages. At the first stage, a preference for a 146 candidate drives a voter’s party choice. At the second stage, we assume a voter has a partisan preference and thus the personal vote drives the allocation of preference votes. Data restrictions prevent a clear distinction between first and second stage personal voting in Indonesia. My underlying argument, however, assumes that higher preference voting indicates higher instances of first-stage personal voting. Explaining the personal vote A legislative candidate can attract support through an emphasis on their actions or attributes, though the relative ‘personalism’ of a campaign varies widely across political systems. To understand why some legislative campaigns are candidate-centered and others party-centred it is necessary to consider how and why candidates make personal appeals.185 The ‘supply’ of personal appeals is affected by a candidate’s motivation to earn a personal vote and the means the candidate has to pursue a personal campaign. Electoral institutions directly link electoral outcomes to successful personal appeals. There have been several attempts to measure the ‘personalism’ of electoral systems (Carey and Shugart 1995; Johnson and Wallack n.d.). In general, plurality systems provide a strong motivation for personal appeals because winners are determined by the individual candidate’s total vote share. Incentives for personal appeals within proportional systems can vary widely depending on their specific features. In open-list electoral systems there is considerable incentive to pursue a personal vote as a candidate’s victory is determined by the preference vote count (Carey and Shugart 1995). This is especially true in high magnitude electoral districts; however, even when lists are closed there can be incentives for personal appeals if district magnitudes is low (Shugart, Valdini, and Suominen 2005). In short, electoral institutions determine the motivation to earn a personal vote. Partisan structures affect the means for pursing a personal appeal. First, centralized candidate selection mechanisms limit a candidate’s ability to carve out a policy niche. Party leaders that hold the ability to remove their co-partisans from the candidate list carry an effective ‘stick’ with which to enforce discipline. Candidates and legislators are more likely to toe the partisan line for fear of antagonizing party leaders 185 I will refer to any attempt to cultivate a personal vote as a ‘personal appeal.’ 147 (Carey and Shugart 1995). Second, parties with centralized systems of campaign finance provide less incentive for candidates to pursue a personal vote, if only because resources to do so are limited (Samuels 1999). A candidate in control of her own resources can run a more rigorous personal campaign.186 Just because a candidate has the means and motivation to pursue a personal appeal does not entail the voters will be receptive. Different electorates have different ‘demand’ for personal appeals. These demands link specific socio-economic conditions to distinct credit-claiming activities or candidate attributes. For instance, it is often held that rural communities have a ‘friends-and-neighbours’ style of politics conducive to personal appeals. Rural areas prefer voting for a candidate they have a personal connection with, typically a candidate with local roots. Urban areas, on the other hand, are thought to be shaped by either ideology or political machines, both of which make politics less personal. Economic conditions can also affect demand for personal appeals. Bribery and other direct forms of material support sit on the edge of ‘credit-claiming’ activity and can take on legal and illegal forms. Voters with a low economic security are more willing to ‘sell’ their political support in order to attain immediate material benefits (Nichter 2008; Stokes 2005). Whether economic conditions lead to personal appeals is dependent upon the institution context; poverty can just as easily produce impersonal machine politics. Despite the institutional caveat, it is plausible that low levels of economic development or sharp levels of economic inequality produce an environment conducive to personalized campaigns based on direct exchange relationships. No existing theory emphasizing either the ‘supply’ or ‘demand’ for personal appeals can explain cross-district outcomes in Indonesia. On the supply-side, personal voting rates tend to decline with increases in district magnitude, the exact opposite of what is suggested by present theory. Electoral districts with high rates of personal voting could be strongholds for parties that offer candidates structural and/or ideological 186 Legislative institutions can affect the ‘means’ by which a personal appeal is made. In decentralized legislatures politicians tend to have more opportunities for to pork-barrel spending or more freedom to carve out a distinctive policy position, whereas legislatures that are tightly controlled by party leaders offer fewer opportunities to carve out distinctive positions. Given that legislative structure is constant across districts it cannot explain within-case variation. 148 incentives to pursue a personal vote. As will be shown in the Indonesian case, however, personal voting rates vary systematically across districts even within political parties, strongly suggesting explanations that go beyond partisan factors. ‘Demand’ explanations do no better. Rural districts tend to have lower rates of personal voting. While some of the districts with the highest personal voting rates were poor, many of the poor districts, especially those in Java, did not have high personal rates. Ethnic diversity and personal voting It is plausible that the ‘demand’ for personal appeals is positively associated with ethnic diversity. Voters tend to prefer electing co-ethnics. As outlined by Chandra, an individual derives some level of psychic satisfaction when a co-ethnic succeeds at the polls (Chandra 2004). Additionally, ethnic representation is important in ensuring a dependable flow of the spoils of office. Elected posts help ensure that an ethnic group will attain a share of particularistic goods, such as jobs or pork-barrel spending, as well as help prevent any policy initiatives that are perceived as potentially harmful to the group’s interests. Ethnic diversity heightens ethnic consciousness. In an ethnically homogenous electorate, voters are more likely to assume their ‘ethnic’ interests are looked after. The same assumption cannot be made in a diverse electorate. The existence of multiple groups increases the saliency of ethnic identities and ‘reminds’ voters of their ethnic interests. Ethnic diversity begets ethnic competition. Ethnic voting can lead to support for a particular party or politician. An ‘ethnic party’ acts as the political vehicle of one particular ethnic group. Ethnic parties communicate their identity either explicitly through their official platform and policies or implicitly through the overwhelming representative presence of one group that dominates the party structure. Even parties that are not explicitly ‘ethnic’ can establish a reputation for defending the interests of a particular ethnic group through policy positioning or incorporating ethnic representatives into the party structure. Ethnic voting at the candidate level occurs when the ethnic background of a candidate brings an electoral boost solely based on her connections with co-ethnics. Dress, name, speech, and even facial hair can all potentially carry vital cues that signal to all voters a candidate’s ethnic affiliation and loyalties. In diverse electorates that lack 149 ethnic parties, candidates are well positioned to pursue a primarily attribuitional personal vote by appealing to their ethnic community. Putting the argument together, we should expect higher rates of preference voting in diverse electoral districts in Indonesia. Because no ethnic parties exist, ethnically conscious voters in diverse districts should support co-ethnic candidates on a party’s list. Looking across districts, there is support for the argument. Some of the highest rates of preference voting do occur in diverse districts. Nonetheless, there are key pieces of the causal story that do not hold. As explained below, preference for co-ethnics is not stronger in diverse districts and there is little evidence candidates consciously campaign using ethnic cues. The correlation between diversity and preference voting is indirect and works through the mechanism of rent-seeking and patronage. Rent opportunities and personal voting Politicians and political parties can earn support through the direct provision of gifts and favours. The terminology used to describe exchange relationships is dependent upon the expectations actors have about the endurance of the relationship. The term bribery refers to the exchange of goods for political support in a onetime interaction, whereas clientelism refers to an enduring exchange relationship between patron and client (Hicken 2011). Below I will use the term clientelistic goods to refer to the material gifts and/or favours a patron provides his/her client in exchange for political support. The provision of material goods before an election allows candidates to rally their base and ingratiate themselves with voters who may be indifferent or mildly opposed to their partisan identity or policy positions. Models of exchange relationships have focused largely on pre-electoral bribery (Nichter 2008; Stokes 2005). Yet paying for support at election time is only one aspect of the exchange relationship. Candidates not only provide gifts at election time, they offer promises of future support (Chandra 2004; Remmer 2007). Jobs, government favours, and other direct forms of post-election assistance are implied in many exchange relationship. Delivery of post-election clientelistic goods serves as a method of repaying supporters and maintaining ties to the base. Promises of clientelistic goods, however, can only be credible when voters anticipate elected officials can manipulate state resources after an election. 150 Vote buying, time horizons, and the personal vote Personal voting behaviour can be partially explained by the opportunities to provide post- election favours. It is useful to think about the sequence of an exchange relationship. In the first place, candidates provide voters with material inducements to increase their chance of winning. Voters, for their part, may agree to sell their support because they value the direct material gain more than they value supporting their preferred political contender. Even so, what prevents a voter from defecting and casting a vote for an alternative political option? By taking a gift and defecting, voters can have their cake and eat it too.187 Knowing this, politicians should not offer gifts and favours.188 Recent research on the issue focuses on monitoring and time horizons (Chandra 2004; Nichter 2008; Stokes 2005). Political actors, especially those in developing democracies, develop mechanisms to monitor voting behaviour. Ostensibly secret ballots can still be discerned. Clever political operators can tamper with counting methods such that voters’ choices are revealed, tweak voting machines to broadcast polling booth decisions, and design ballots in ways making them easy to manipulate. Close monitoring allows political actors to sort out who has been loyal and who has reneged on their past promises. Revealed defection only becomes a problem for voters if they expect consequences. In most democracies, however, elections are repeated interactions. Stokes, for instance, argues parties are less likely to bribe known defectors in future elections (Stokes 2005). In her model, receiving a bribe from Party A and supporting Party B can earn a voter a reputation for defection, which costs the possibility of receiving a bribe from a political contender. The costs of defection go beyond simply forfeiting the possibility of receiving a bribe during a future electoral cycle though. A reputation for 187 The gastronomic metaphor dovetails with actual education campaigns encouraging voter defection. Shaffer provides a list of examples: Many civic educators around the world have kept the message simple and palatable to voters: accept the money, but vote your conscience. In Bulgaria, the party representing the Roma told their supporters to “eat their meatballs but vote with your heart.” Civil society groups in Zambia urged voters to “eat widely but vote wisely.” Jamie Cardinal Sin, archbishop of Manila during the twilight years of Ferdinand Marcos, advised voters to “take the bait but not the hook.” (Shaffer 2007, 161) 188 Stokes models this interaction as a prisoner’s dilemma, finding that competitors should not bribe in the absence of monitoring and repeated interaction (Stokes 2005, 319-320). 151 defection can also cost a voter between elections, and I argue below that these perceived costs can vary by district. Rent opportunities and clientelist exchange An elected legislator’s ability to reward supporters and punish defectors depends on the resources at hand after the election, which I refer to in this dissertation as rent opportunities. In electoral districts with plenty of rent opportunities, the probability of receiving a government favour from a politician is comparatively high. The opposite is true in areas with fewer government jobs and stronger anti-corruption norms. The candidate in the high rent area, then, can credibly commit to providing clientelistic goods (and withholding clientelistic goods from defectors), whereas the candidate in the low rent area has a problem making the same promise. Given that the cost of defection is high in some districts and low in others, we should see different patterns of campaigning and voting depending on the social context. When plenty of rent opportunities exist, voters have an incentive to loyally attach themselves to a patron. Knowing this, candidates use gifts and favours to build a relationship with voting blocs. This pattern is reversed in areas with fewer rent opportunities. Where the future costs of defection are low, voters have little incentive to stay loyal and candidates have little incentive to provide clientelistic goods. This leads to Hypothesis 1: H1: A high percentage of voters casting optional preference votes is positively associated with a high level of rent opportunities The Hypothesis put forward is in line with previous research on preference voting in Italy that finds preference voting highest in areas with a high degree of “traditionalism” and/or low social capital (Katz and Bardi, 1980; Putnam, 1994). The focus of the Italian case work, then, has been on the social ‘constraints’ (or lax thereof) placed on a politician with regard to the direct distribution of state resources. I build on this research by narrowing in on the relationship between preference voting and the stock of sub-national resources that can be plausibly exchanged for votes. 152 Case background: campaigns in Indonesia Running a political campaign in Indonesia is a costly pursuit. The minimum an active candidate can expect to pay is 50,000,000 Rupiah, approximately $5,000 US at the time of the 2009 campaign. To place this in perspective, GDP per capita in the same year was $2,349 US. And this is the cost that a modest candidate running under a marginal party label can be expected to pay for a municipal-level position. Tweaking key variables such as party prominence, vigour of the campaign, and the level of governance produce higher campaign costs. An active campaign under a major party label for a seat in the DPR can cost over a 1,000,000,000 Rupiah ($100,000 US). In short, Indonesian campaigns are remarkably expensive affairs. Candidates use multiple campaign methods to get their message out. A considerable portion of a candidate’s funds are directed towards advertising. During the run-up to the election, Indonesia’s streets are lined with colourful candidate campaign posters (‘baliho’). Newspapers are filled with candidate advertisements. Members of a candidate’s Success Team (‘Tim Sukses’) pass out name cards and calendars. All of these means of direct advertising convey a candidate’s basic information: the position they are running for, party label they are running under, their position on the candidate list, and a brief slogan.189 In many cases a picture of the party leader or another prominent individual will also be displayed. The majority of advertising is financed by only one candidate; however, a sizable minority promote multiple candidates from the same party running for different levels of governance. Advertising is expensive but it is not the only form of campaigning. Gift giving is also ubiquitous. Among other things, candidates set up stalls (‘warung’) providing discounted food, they repair community infrastructure, and they supply simple consumer goods. As one witty observer noted, Indonesian campaigns centre on the love of MARKOS, an acronym covering food (‘makan’), cigarettes (‘rokok’) and assorted daily necessities (‘sembako’).190 189 For examples, see Figures 18 – 24. 190 ‘Sembako’ is itself an acronym for the nine necessities of daily life, including cooking oil, rice, fuel, and other simple commodities. 153 While gift giving can take many forms, it is often geographically targeted. By design, some of the larger projects target specific areas. Candidates know where they have repaired religious buildings and set up food stalls. Even the smaller gift-giving often follows a geographic logic. When asked to describe their constituencies, candidates typically mentioned both a diffuse support base (‘women’, ‘farmers’) and a more geographically concentrated base, such as a set of villages or an area with many relatives. Practical reasons make it easier to focus on the geographic base. For instance, one DPR- RI candidate in the Medan area mentioned a support base amongst ‘youth’, a local religious network, and neighbourhoods dominated by ethnic Bugis migrants.191 Upon completion of the interview, the candidate and members of his team loaded flats of bottled water into a sport utility vehicle and took off to campaign in a few select outlying villages. The candidate in question did have strong network connections with his party’s youth organization, but when it came time to pass out simple consumer products he focused on promising geographic constituencies. Monitoring and punishment In addition to advertising and gift giving, a third major campaign cost candidates face is poll monitoring. Each candidate employs a team of poll monitors, known as witnesses (‘saksi’). The size of the team varies by the candidate’s resources. Witnesses, most often young men, are typically provided with some compensation for their efforts.192 Their job involves monitoring the process and tabulating the results at the various polling stations. In comparison with Western nations, the process of voting and vote tabulation is a remarkably public event. On voting day all business ceases and otherwise busy roads empty. A massive number of polling stations service small, often close-knit communities.193 Polling stations have official observer sections where paid operatives and other monitors watch and record the proceedings. Knots of people gather to enjoy the spectacle. Names are read off the voters list in a pre-arranged order. Called voters line up 191 The candidate was not himself from that ethnic group. 192 Saksi also hope their efforts will put them first in line for distribution of clientelistic goods. During a candidate’s organizational meeting on the eve of the 2009 election, the assembled young men openly shared with me that they hoped their service would land them post-election employment. 193 In 2009, electoral authorities put the number of polling stations at 519,920. Thus there was approximately 1 station for every 200 voters. 154 and wait their turn in the booth. Following the completion of voting, election officials publically tabulate the results at each station. On an individual level, each voter’s ballot choice remains private. However, the small-scale natures of tabulation processes allow candidates to monitor their campaign investments. Witnesses are able to gather several important pieces of information. First, they have an idea of who showed up to cast a ballot and who chose to abstain. Candidates, then, have an idea of turnout both at the local and individual level. Second, candidates know which areas they were strong in. By the end of the process, candidates are knowledgeable of their support base. They know who hustled for them as members of their team success. Their close associates know which becek drivers and warung operators displayed the candidate’s campaign paraphernalia. They know which local communities offered their support. And, if they are competent, they know which communities ‘defected’ by accepting their gifts and voting for their opponents. The consequences of defection and loyalty are difficult to pinpoint. We know that legislators hand out favours between elections. These favours can take the form of the manipulation of state policy, such as the provision of a job, or simply providing direct financial help, the cost of which the politician must latter recoup through corruption or influence peddling. Supporters expect to be rewarded. Nonetheless, systemic data on this is lacking. Anecdotes suggest punishment does occur. One such story was related to the author in Pematang Siantar, North Sumatra. In this story a candidate from the 2004 election (level of government unmentioned) helped fund the repair of water facilities within a small community. Despite this support, the community in question returned a dismal number of votes for the candidate. Feeling spurned, the candidate returned and took back the equipment he had previously gifted to the community. The particular story itself is potentially apocryphal and told mainly for humorous effect. Similar stories appear in the press though. In one incident a candidate returned to a village in West Java to request the return of his 50,000 Rupiah (approximately $5 US) gifts. Villagers had accepted his gifts prior to the election, yet the candidate managed to 155 obtain only one vote from the area (Jakarta Post April 16, 2009). 194 These stories hint at several processes that are taken as given. First, material inducements are geographically concentrated. Second, campaign investments are monitored. Third, gifts do not guarantee support. Fourth, spurned politicians carry a grudge.195 Decentralization and clientelism Prior to the 2004, Indonesia enacted a series of decentralization laws that substantially expanded the political, fiscal, and administrative autonomy of sub-national units. Whereas during the Suharto years the national executive had leeway to pick and choose provincial governors (and, by extension, municipal executives), after decentralization local executives were made accountable to newly empowered local legislatures. Control over many local rent opportunities (jobs, transfers, etc) shifted from the national to the local level. This chapter, however, focuses on national level voting trends. Which raises the question: why would voters expect national politicians to pass out sub-national goods? Two potential answers exist. First, voters may simply not know how to match up levels of governance with jurisdiction over things they want. National level politicians may in fact have little power of local state activity. Nonetheless, voters expect national politicians to hand out local goods because they are unsure of the new jurisdictional boundaries. A second answer suggests that voters may correctly surmise that national level politicians retain significant influence and expect their national patron to intervene in their favour. On the one hand, there are signs that national legislators use their influence to directly channel regional transfer funds to preferred projects and uses. Perhaps more importantly, national level candidates frequently exist as part of a broader network that connects all the way down to the sub-national level. Indeed, given the expense of 194 In a less direct, though more tragic, form of punishment one legislative candidate took her life after recording a dismal vote within her own village (Jakarta Post April 15, 2009). In the immediate aftermath of elections mental health facilities swell with an intake of depressed candidates. According to one mental health professional, “several patients frequently talked in their sleep, asking for their money back because of their failure to gain a significant number of votes” (Jakarta Post April 16, 2009). 195 The story carries one additional sub-text that falls outside of the prior theoretical discussion; namely, even losing candidates may have the means and motivation to punish defectors. This is a particularly important point in Indonesia, where the wealthy citizens who run for office tend to be the same wealthy citizens the poor turn to for jobs, loans and other forms of assistance. 156 national-level candidacies, national level politicians are often among the most prominent members of a network. Take, for instance, Mudaffar Syah (aka the Sultan of Ternate). In 2004 he ran as a national candidate in North Maluku for PDK, a minor upstart. His party may not have been destined for national greatness, but voters knew that a vote for the Sultan and PDK would connect them to a network that would control substantial local resources. Likewise, Maluku candidate Mirati Dewaningsih’s strength in 2009 was certainly not because of her affiliation with the locally insignificant PKB. Rather, voters understood that her marriage to the local Bupati (municipal executive) meant that a vote for her was vote for her husband’s broader network. The vertical integration of political networks is strengthened by the simultaneous legislative election cycle. This study will not resolve which of the two mechanisms – voter confusion or multi-level influence – is at play in Indonesia. Both are plausible. The most pertinent point is that voter behavior in electoral districts with a high level of rent opportunities is conditioned by sub-national dependence on the state sector. The political-economic environment leads to distinct patterns of interaction between voters and candidates. Theory testing: personal voting across electoral districts Dependent variable: personal voting rates Having described the Indonesian context, I move on to statistical testing of my hypothesis. My dependent variable is the percentage of voters who cast a preference vote by party by district. An example helps illustrate. In District 1, Party A receives 5000 party symbol votes. Each of the three candidates on the party list receives 1000 preference votes. Thus in District 1, 60% (3000/5000) of voters cast a preference vote for Party A in District 1. The percentage serves as the key dependent variable. Data for the 2004 Indonesian election was gathered from the website of the Indonesian electoral authorities and checked against Kevin Evan’s Pemilu Asia website. I focus on the national elections primarily because the data is available. There are 1656 observations, one for each party in all 69 electoral districts. 157 Independent variable: local rent opportunities Measuring the rent opportunities concept is tricky. Both the prevailing constraints on behavior and the resources available for manipulation remain difficult to pin down. In this chapter, I focus only on the resources available to the sitting (or prospective) legislator. This decision is due to data limitations. While some areas of Indonesia have reputations for being less corrupt than others, it is easier to attain measures of budgets and bureaucratic size than perceptions of corruption. Despite the fact that social norms and law enforcement practices do vary across electoral districts, it is safe to assume that, in general, a low level of constraints exist throughout the country. Even Indonesia’s cleanest regional governments would be considered systematically corrupt by the standards of the established developed democracies. Given the weakness of constraints, measurements that capture the stock of local resources for manipulation serve as a workable proxy of the rent opportunities concept. To measure rent opportunities, I focus on the relative proportion of civil service jobs in the modern sector, defined here as the non-agricultural economy. The use of this measure follows similar practice in both the comparative and Indonesian literatures on patronage (Chandra 2004; van Klinken 2007). A high proportion of modern sector state employees indicates the extent of state involvement in the overall economy. In Indonesia, some provinces like Central Java and Bali have few state jobs. In outlying provinces such as Papua and Bengkulu the state is a very significant employer. Data on employment sectors is drawn from the Indonesian Statistical Yearbook series. I use two different years: 2005 and 1999. The year 2005 is chosen due to its proximity to the 2004 election. While the 2005 data reflect civil service size slightly after the 2004 elections, it is unlikely that the result of the election itself caused any change in the variable in the short period between election and data collection. The 1999 data, generated around the time of the democratic transition, provides a further test of my claim that the patterns of state employment shaping political competition pre-dated democracy in Indonesia.196 196 As an alternative to civil service size, I also test all models using per capita central government transfer as a measure of rent opportunities. Results using the alternative measure appear in the appendices. For more on the construction of the variable itself, see Appendix F, Section 1. 158 Provincial aggregates visually demonstrate the striking correlation between civil service size and preference voting. Figure 17 presents a two-way with public sector size (2005) on the X-axis and preference voting on the Y-axis. There is a strong, positive relationship between public sector size and preference voting rates. The independent variable alone explains just over 50% of the variation in aggregate preference voting rates across provinces. Figure 17 – Preference voting and rents Control variables I include a range of control variables to the statistical model. First, I add district magnitude. Carey and Shugart (1995) suggest that personal vote-seeking incentives in open-list systems increase as district magnitude increases. In the 2004 elections in Indonesia, district magnitude ranged from 3-12, with the size being dependent on both the size of the population and the legacy of traditional political boundaries. To account for economic conditions I add a variable capturing poverty rates. Voters with low economic security are more willing to ‘sell’ their political support in order to attain immediate material benefits (Nichter 2008; Stokes 2005). The benefit of having a powerful patron that can dispense favours is likely of more importance to the destitute. As such, high levels of relative poverty should be associated with higher levels 159 of personal voting. I use data from Suryadinata et al. that captures poverty by municipality which I then use to reconstruct district-level poverty for 2004 (Suryadinata, Ananta, and Arifin 2004). An additional social variable consider is the level of urbanization. There are two potential stories that can be told. The ‘friends and neighbours’ style of politics associated with rural districts could increase the importance of personal appeals. As mentioned in Chapter 4, the interaction between candidates and voters in rural settings often includes provision of gifts and community goods, which could foster candidate-voter bonds. Alternatively, media markets in urban centres could provide voters with increased information about candidates, leading to higher levels of personal voting. Urbanization data comes from Indonesian Electoral Behaviour (Suryadinata, Ananta, and Arifin 2004). Ethnic diversity could produce higher rates of preference voting regardless of the existence of rents. If competitive dynamics in diverse societies lead voters to disregard policy concerns, voters may attach themselves to friendly co-ethnic candidates. Such a dynamic could be particularly pronounced in a country that effectively bars ethnic parties. To control for ethnic context, I used 2000 census data to construct a 0-1 ethnic fractionalization measure for each electoral district. Three party-level variables are also included. The first accounts for the number of candidates on a party’s slate. Higher numbers of candidates mean a higher number of personalized campaigns and more opportunities to forge candidate-voters connections. To construct the measure I divide the raw number of candidates run on a list by the district magnitude. Given that district magnitude tends to follow population levels, the measure roughly captured the number of candidates to voters. When the measure is low, the party has few local agents pursuing preferences votes; when the measure is high there are a large number of candidates appealing for support relative to the population. The second party-level variable accounts for the potential effect of gender on personal voting. There is some evidence that women candidates are at a disadvantage when competing for intra-party preference votes (Valdini 2006). Low-information voters use gender as a short-cut to infer qualities about a candidate. Popularly held stereotypes about men are associated with characteristics that people desire in a politician (e.g. courageous, rational, strong) while female cues tend to be associated with less desirable 160 traits (e.g. emotional, frivolous, weak). When a large percentage of a party’s candidates are women, voters could feel less connected to an individual candidate on the party’s list, despite being favourably disposed to the party’s message. Thus a high proportion of women candidates could produce a lower personal voting rate. To construct this measure I simply divide the number of women by the overall number of candidates. The third party-level variable takes into account the existence of local roots. Candidates with local roots tend to perform better than parachute candidates in electoral systems with preference voting (Shugart, Valdini, and Suominen 2005; Tavits 2010). The logic of the variable, then, is similar to the gender variable above: if a list has more local candidates, it should have a higher aggregate personal vote. To measure the concept of local roots I rely on a candidates stated place of residence. Candidates are coded as ‘local’ if they resided in the district they sought to represent. To construct the ‘local’ measure, the number of local candidates is divided by the total number of candidates. Results I run the model using a simple OLS regression with standard errors clustered by party. Parties vary in nomination procedures, resources, ideology, and national campaign style. While I expect all parties to be affected by the variables discussed, the effects are likely mediated by partisan factors. Assuming that standard errors are correlated within parties is a more plausible position than assuming the existence of complete independence for each branch of each party.197 Table 12 contains results. Models 1 and 2 include a different operationalization of the key rent opportunities variable. In both models, the rent opportunities variable is positive and statistically significant at the P < 0.01 level. An increase of modern sector civil service size by 1% correlates with a just over a 1% increase in preference voting rates. Using Clarify to generate predictions, an increase of one standard deviation from the mean produces a 6.9% increase in preference voting using the 2005 data and a 6.1% increase using the 1999 data. 197 I ran the models using fixed effects by party. The alternative model specification does not alter the results. See Appendix F, Section 2. 161 Table 12 – Rent opportunities and preference voting Rent Opportunities and Preference Voting (OLS) Model 1 Model 2 Estimate (std. err) Estimate (std. err) Civil Service Size (2005) 1.04*** (0.07) Civil Service Size (1999) 1.07*** (0.06) District Magnitude 0.74*** (0.16) 0.59*** (0.16) Urbanization 0.09*** (0.02) 0.07*** (0.02) Poverty -0.04* (0.02) -0.02 (0.02) Ethnic Fractionalization 5.51*** (0.95) 6.85*** (1.03) Candidate List Size 0.21*** (0.02) 0.21*** (0.02) Home Candidates 0.04** (0.02) 0.05** (0.02) Women Candidates 0.005 (0.02) 0.001 (0.02) Constant 10.91*** (3.06) 11.35*** (3.05) Observations 1646 1646 R2 0.5110 0.5118 *p < .10 **p < .05 *** p <.01 The magnitude of the rents variable can be demonstrated through statistical simulations. Using Model 1 as a baseline, I generate predicted values on the dependent variable while adjusting the public sector size. Table 13 displays the results. Five predictions are presented, with public sector size set to actual minimum, maximum, and quartile values. All other variables are set to the mean. Quartile values on civil service size are presented along with the corresponding province that matches that value. Predicted values indicate that moving from the minimum to the maximum public sector size results in a 60% higher preference voting rate. 162 Table 13 – Predicated impact of rents on preference voting Predicted Impact of Rent Opportunities on Preference Voting Rent Opportunity Level Public Sector Size (% Civil Servants) Province Name Predicted Preference Voting (% Voters) Minimum 3.2% Banten 39.4 Lower Quartile 7.8% Nusa Tenggara Barat 44.1 Median 11.2% West Kalimantan 47.7 Upper Quartile 17.7% Aceh 54.6 Maximum 29.1% Papua 66.5 Returning to Table 12, the district-level control variables are all statistically significant, though the poverty variable slips below significance in Model 2 and changes signs from Models 1/2 to Model 3. District magnitude is signed in the expected positive direction and significant at the .01 level. An increase of district magnitude by one seat correlates with a 0.7% increase in preference voting. Urbanization correlates with increased rates of preference voting, though the effect is modest. An increase of one standard deviation in urbanization correlates with a 2% increase in preference voting. This result is strong but unexpected. I was told by candidates that spending on gifts tends to increase in rural areas, and my exchange based theory would suggest this should in turn increase personal voting. Clearly, the data does not support this hypothesis. There are several potential interpretations of the correlation between urbanization and preference voting. First, the results could represent the relative political awareness of urban voters. With higher levels of education, increased access to media, and an abundance of candidates competing for their vote, urban voters may have had an easier time familiarizing themselves with the new system and selecting a suitable candidate. Second, the results could indicate the presence of collective decision-making at the village level in rural areas. Feith’s account of the 1955 Indonesian election emphasizes the pivotal role played by village notables in determining the voting behaviour of fellow villagers (Feith 1957, 21-37). It may be difficult for candidates to build a strong candidate-voter bond when village notables act as intermediaries between candidates and voters. Whatever the case may be, the strength of the urbanization effect only presents itself once rent opportunities have been controlled for. Thus the finding is intriguing but does not present a serious challenge to my causal story. 163 Surprisingly, poverty levels are correlated with lower personal voting rates in Model 1, though the effect is slight; an increase of one standard deviation in the poverty level correlates decreases preference voting by 0.5%.There also tends to be more preference voting in ethnically diverse electoral districts. A one standard deviation increase in ethnic fractionalization produces a predicted 1.9% increase in preference voting. Two of the three party-level variables – candidate list size and percentage of local roots - consistently correlate with higher preference voting rates. Candidate list size has the largest impact on personal voting rates. In a district magnitude of 10, increasing the number of candidates on a party’s list from 5 to 10 would be expected to increase personal voting rates 10.5%. This is a substantial change, and demonstrates that exposure to candidate campaigns increases the propensity to cast a preference vote. Local roots have a small but statistically significant impact. A list where 50% of the candidates have local roots is predicted to have 2.1% fewer preference votes than a list where 100% of the candidates have local roots. On the other hand, the relative proportion of women candidates has no effect on preference voting rates. To check the robustness of the results I test for the possibility that results are driven by party type, population size, or separatism. First, the results could plausibly be driven by the dominance of particular parties in particular areas. I re-ran Model 1 for each individual party.198 In every party the rents variable reached statistical significance of at least P<.05; in fact, it is the only variable to consistently reach statistical significance across parties. Coefficient sizes range from a low of 0.7 (Partai Buruh Sosial Demokrat, PBSD) to a high of 1.7 (Partai Penegak Demokrasi Indonesia, PPDI), such that one percentage increase in civil service size increased preference voting rates in by 0.7% in PBSD and 1.7% in PPDI. There is a noticeable trend in which the nationally small ‘chicken flea’ (gurem) parties tend to be those that are most affected by rent opportunities. The gurem parties have relatively low average personal voting rates, a consequence that stems from the low number of candidates these parties run in most districts. They have organizational and electoral breakthroughs in high rent areas. Despite the fact that the effect of the rents variable was largest for the gurem, it also had a 198 See Appendix F, Section 3. 164 significant impact on the larger and more established parties. We can safely conclude that the relationship between rent opportunities and preference voting was not driven by a small number of factionalized parties performing well in particular areas. Second, I investigate the possibility that an omitted variable may be shaping both rent opportunities and preference voting. This raises a key question: why do rent opportunities vary? The simple answer is that rent opportunities tend to correlate with population size. The centre-periphery conflicts and bargains that characterized the Indonesian state building process produced small sub-national units in certain areas. In contemporary Indonesia, small units receive more per capita transfers and thus have a relatively large pool of resources for politicians to manipulate. However, small population size could also facilitate a ‘friends-and-‘neighbours’ style of politics that brings candidates into direct contact with constituents, thereby increasing the propensity to preference vote. It is thus possible that small population size rather than resource availability drives the correlation between transfers and preference votes. To check for the possibility of a population effect, I add several permutations of population size to the statistical model. The first capture population of the electoral districts, the second captured the population of the province. Neither of the population variables have a consistent relationship with preference voting. More importantly, the addition of the variables did not substantively alter the relationship between the rents measures and preference voting.199 Third, I investigate the potential effect of separatism. In order to reward local allies and mitigate local grievances, the Indonesian state transfers relatively large sums of money to regions with separatist movements. The problematic centre-periphery relationship that affects state policy also affects internal party dynamics. Across the country, favored local candidate tend to receive undesirable list positions. Yet the sting of Jakartan insensitivity is keenly felt in restive regions. It is possible that voters in regions with active separatist movements are more motivated to cast a preference vote in order to support a regional voice and spite the national parties. To test the possibility that separatism drove results, I drop electoral districts in regions with separatist movements. In the lead up to the 2004 elections, the regions with 199 See Appendix F, Section 4. 165 active separatist movements were Aceh and Papua.200 Dropping the separatist districts from the sample, however, does not substantially change the correlation between the rents variable and preference voting.201 While voters in separatist areas may use their preference vote to register their protest with the national system, this dynamic itself cannot explain the relationship between rents and preference voting. Alternative explanation: ethnic diversity and personal voting Preference voting rates correlate closely with rent opportunities, defined here as the relative size of the public sector. Sub-national rent opportunities, however, are shaped by ethnic structures: politicians in ethnically diverse regional governments have greater access to patronage than their counterparts in ethnically homogenous regions. Despite the results thus far, ethnic diversity may still independently produce higher preference voting rates. The communal voting model has distinct observable implications as they relate to campaigns, opinions, and voting behaviour. First, during campaigns, there should be a relative propensity to use ethnic symbolism. Where voters are pre-disposed to consider ethnic backgrounds, candidates play up to these ethnic biases in order to build a support base. If the communal voting story is correct, then, campaigns in diverse electoral districts should focus not only on candidate personalities but also their ethnic background. Second, if voters in diverse areas are primed to consider a candidate’s ethnic background, they should report opinions that reflect these ethnic preferences. A propensity to express preferences for co-ethnic candidates should be observable in ethnically diverse electoral districts. 200 The criterion of ‘active’ excludes the borderline cases of Riau and Maluku. Riau’s post-Suharto separatist movement was always more of an intellectual and cultural enterprise and was effectively moribund by 2004. In Maluku, the Republik Maluku Selatan (Republic of South Maluku, RMS), which had been first defeated in the 1950s, could not even muster a political comeback e during the province’s 2000 communal conflicts. In Papua, on the other hand, the widely respected Presidium Dewan Papua (Papua Presidium Council) took up the separatist torch from the flagging Organisasi Papua Merdeka (Free Papua Movement, OPM), keeping the demand for independence at the political forefront in the long restive region. In Aceh, the crackdown on non-violent calls for independence in the late 1990s precipitated fighting between government troops Gerekan Aceh Merdeka (Free Aceh Movement, GAM) rebels which continued until the 2004 tsunami finally created the conditions for a peace agreement. 201 See Appendix F, Section 5. 166 Ethnic diversity and candidate campaigns I examined the campaigning implication of the communal voting argument during field research. In order to test the hypothesis I leveraged intra-district variations in ethnic structure. During the 2009 election I visited two distinct areas in the electoral district of North Sumatra III: Karo and Pematang Siantar. North Sumatra III is a very diverse electoral district. People come from a range of groups, the largest being Toba and Karo Batak, Javanese, and Melayu. Kabupaten Karo has a relatively homogenous population made up of Karo Batak. Kota Pematang Siantar, on the other hand, has a mixed population. Where Karo is something of an ethic enclave, Pematang Siantar is a rough microcosm of North Sumatra’s diverse ethnic make-up. The strictest reading of the communal voting hypothesis would suggest that all voters in North Sumatra should be primed to consider ethnicity. Candidate campaigns should cue on ethnicity frequently and there should be no intra-district variation. A looser reading suggests candidates in Pematang Siantar should cue on ethnicity more than their counterparts in Karo. If regular daily contact breeds ethnic consciousness and ethnic competition, then Pematang Siantar should exhibit more outward signs of ethnic politics. A finding that no ethnic cueing takes place or more takes place in Karo would challenge the communal voting hypothesis. I conducted interviews with candidates to probe their campaign strategies and their mental mapping of their district. Interviews were conducted in late February-early March 2009. Within Karo, interviews were limited to the regency capital of Kabanjahe. Due to ease of access, I spoke primarily to candidates running for sub-national offices (municipal and provincial). Observation and interviews in both locations did not find any evidence to support the communal voting story. In general, candidates did not report thinking of voting blocs in clearly ethnic terms. With the respondents I spoke with, familial and religious ties were the most frequently mentioned network connections used when targeting a support base. While religious and familial networks tend to follow ethnic lines, it is notable that ethnicity was not a readily mentioned criteria for delineating support bases. Though interviews exposed little intra-district variation in the use of ethnic campaigning, observational data revealed distinct differences. Within Karo, ethnic cueing 167 in promotional material was the norm. Candidates proudly flaunted their connections to the community through slogans and dress. Figure 18 presents a typical example. In the photo, Ngasup Karo-Karo Sitepu, a PKB candidate for national office, covers the basics of the ethnic cue: his name (Karo-Karo) quickly reveals his background, his business suit is adorned with traditional Karo garb, and one of his primary slogans contains a popular blessing in Karo dialect (‘Mejuah-juah’). Although my example includes clear ethnic cues by a candidate at the national level, the tendency was also present for those running for sub-national office. Indeed, the tendency was so strong that even non-Karo candidates could be found using Karo cues. Figure 18 – DPR candidate in Karo In contrast to the obvious displays of ethnic affiliation found in Karo, candidates in Pematang Siantar were relatively cautious in their ethnic cues. Whereas the norm in Karo was to deliberately emphasize ethnic identity through clothing or slogans, candidates in Pematang Siantar tended toward an ethnically neutral presentation. Edward Hutabarat’s poster, presented in Figure 19, is emblematic. Local voters could tell by Hutabarat’s name that he is descended from Toba Batak parents. But he did not promote this identity. The symbolism was largely nationalist; his slogan emphasized his 168 commitment to education and Indonesia’s constitutional legacy. Although the Toba constitute a plurality within Pematang Siantar, Hutabarat made no deliberate effort to appeal to his ethnic roots. Figure 19 – DPR candidate in Pemantang Siantar The disjuncture between symbolism in Karo and Pemantang Siantar is unlikely to be random. Decisions to cue or not to cue on ethnicity are deliberately made. Candidates within large electoral districts will even make a range of promotional material so as to fine tune their symbolism. Figure 20 and Figure 21 contain two of R.K. Sembiring’s posters. In Karo, Sembiring could be found donning traditional clothing, emphasizing his connections with the Karo ethnic group. When promoting himself in Pematang Siantar, however, Sembiring replaced the Karo head-dress with the more neutral Peci. In Figure 22 and Figure 23, Yopie Batubara demonstrates a similar strategy. His Karo poster boasts his ceremonial adoption by the Tarigan, a prominent local clan, while his advertisement in more cosmopolitan Medan downplays his Karo ties. 169 Figure 20 – DPR candidate Sembiring in Karo Figure 21 - Sembiring in Pematang Siantar Figure 22 – DPD candidate Batubara in Karo Figure 23 – Batubara in Medan 170 Candidates cue when they think there is a strong communal attachment to symbols shared by most voters. Even for regency-level elections in Karo, where the large majority of voters and candidates came from the same ethnic group, ethnic cues were prevalent. This pattern is not limited to North Sumatra. In Figure 24, Balinese regency- level candidate I Made Sumer takes this style of ‘playing to the base’ ethnic politics to the extreme. Sumer dispensed with the standard photo and slogan, and provided in their place merely an advertisement for a local Balinese arts festival. The point of Sumer’s poster did not to set him apart from other candidates with different backgrounds; in fact, his district was largely Balinese and most candidates used Balinese symbolism in the campaigns. Rather, Sumer’s poster underlined his attachment to shared communal values. Figure 24 – DPRD candidate Sumer in Bali This preliminary analysis indicates that daily proximity and regular group conflict did not drive deliberate ethnic campaign cues. When it comes to promotional material, diversity breeds caution in Indonesia (at least in legislative elections). This could be because candidates perceive an electoral cost attached to an ethnic appeal. Or it could be that societal norms for ethnic tolerance simply restrain behaviour regardless of the potential electoral effect. It is not obvious, however, that candidates cue because they 171 want to distinguish themselves from their competitors. Ethnic appeals appear to be most strident in homogenous areas, the places we might think the ethnic factor to be least salient. Ethnicity and public opinion A second implication of the communal voting model is higher ethnic consciousness among voters in ethnically diverse electoral districts. Where ethnic groups are in competition, there is an increased saliency of ethnic identity markers. This increase in ethnic consciousness could account for increased rates of preference voting. To explore the potential effect of ethnicity on political opinions I rely on an opinion survey conducted by the Centre for Strategic and International Studies (CSIS), a Jakarta based think-tank.202 The survey was conducted from May to June in 2008, during the run up to the legislative elections. Answers reflected opinions prior to full scale candidate campaigns. A total of 3000 Indonesian’s were polled. Polling only covered 13 of Indonesia’s 33 provinces; however, the districts were selected to reflect social and regional trends throughout the country. Thus the sample includes provinces from Java, Sulawesi, Kalimantan, Sumatra, as well as the Lesser Sundas. There was no clear question asking opinions regarding ethnic preferences for legislative candidates. The closest proximity asked respondents whether they prefer to choose a presidential candidate from their own ethnic group. Preferences for president are an imperfect replacement for legislator preferences. Given that a strong plurality of Indonesia’s electorate is Javanese, the presidency has been dominated by politicians with an ethnic Javanese background. All but one of Indonesia’s past presidents has been from this group and most major presidential candidates also hail from Java.203 Thus non- Javanese voters may be conditioned to accept Javanese presidential candidates, whereas they would be more strident in their demand for legislators from their own group. Despite the problems with the question, it still clearly asked about ethnic preferences and thus may be used as a rough measure of ethnic consciousness. 202 Thank you to Sunny Tanuwidjaja for generously providing the data. 203 Habibie, who served as president during the post-Suharto transition period, was the one exception to Javanese dominance. He was popularly identified as an ethnic Bugi, though his mother was Javanese. 172 The dependent variable captures the existence of stated ethnic preferences. The initial question asked “From the criteria listed here, I prefer to choose a president in 2009: From the same ethnic group as myself.” Three options were listed: yes, no, and do not know. I construct a binary with those answering “yes” coded as “1”, and all others coded as “0.” By separating those respondents who clearly stated an ethnic preference from those who were ethnically neutral or unsure of their preferences, the variable captures the most active and self-conscious communal minded voters.204 The key independent variable measures ethnic diversity by electoral district. If ethnic diversity produces ethnic competition and consciousness, this should be reflected in increased preferences for candidates from the same ethnic group. Thus, once accounting for controls, diversity should positively correlate with stated ethnic preferences. Data on ethnicity was taken from the 2000 census.205 I use the ethnic fractionalization of each electoral district as a measure of ethnic diversity. I add several additional variables. The first individual level variable captured potential gender differences in ethnic preferences. Holmsten et al. argue that appeals for ethnic group representation reinforce patriarchal structures embedded in traditional cultures (Holmsten, Moser, and Slosar 2009). The authors do in fact discover a tendency for ethnic parties to exclude women. Because men tend to be the benefactors of ethnic politics, we might expect them to be more likely to hold strong ethnic biases. Thus Male should be positively correlated with expressed preferences for co-ethnics. A second variable captures age. Indonesia is a young nation. Before the formation of the country, Indonesian’s were comparatively more attached to local identities. As compared to younger, primarily Bahasa speaking Indonesians, the older generation may have more knowledge of local dialects and culture. Knowledge of past customs could increase ethnic consciousness. Thus we might expect older Indonesians to express political preferences for co-ethnics. To measure age I use the 1-3 age variable constructed by CSIS. 204 See Appendix F, Section 6 for information on the coding of all variables in the CSIS dataset 205 Ethnicity data from the census series is broken down by province. All data was collected from BPS library in Jakarta. 173 Education should also influence opinions. Indonesians with higher education levels are more likely to be work in diverse environments and be exposed to cosmopolitan ideas. Educated voters are also more likely to be interested in policy and ideology as compared to personal characteristics. Higher education levels should decrease preferences for co-ethnics. To measure education level I use the 1-5 education variable constructed by CSIS.206 I add a dummy capturing whether or not a respondent is Javanese. Presidential candidates are typically Javanese, and non-Javanese voters are rarely presented with a viable non-Javanese presidential candidate. Given that the dependent variable queries opinions on the ethnicity of presidential candidates, Javanese should express higher levels of support for co-ethnics simply because they have the opportunity to actually support a Javanese candidate. The last individual level variable measures the effect of Islam. Within Indonesia, Islam has been a force for national integration amongst Muslims. The first nationalist mass organization, Sarikat Islam, was founded by practising Muslims. The banner of Islam has also been useful for ambitious non-Javanese office seekers, who have used religious appeals and symbols to expand their influence beyond their relatively small ethnic group. Islam itself was often spread by Malay and/or Arabic speaking spiritual leaders who downplayed local traditions and identities while promoting the broader Islamic identity. Christianity, on the other hand, was frequently spread in conjunction with the promotion of local identities and local dialects (Aragon 2000). Consequently, we might expect Muslim respondents to be less conscious of their ethnic identity and less likely to express preference for co-ethnics. In addition to the individual level controls, I include a variable indicating whether the respondent lives in an urban area. Urban residents in Indonesia have traditionally held cosmopolitan political values. They are more likely to be exposed to the national mass media and tend to be more open to programmatic political appeals. We should then expect urban residents to be less likely to express preference for co-ethnics. To measure 206 Education level also serves as a rough stand-in for economic status. Approximately half of respondents did not provide their monthly income, restricting statistical testing. Conversations with the administrator of the survey reveal that many respondents were simply unable to approximate their monthly earnings because they did not keep track of their various sources of income. 174 whether a respondent lives in an urban area I rely on the CSIS binary city/village variable. I tested my hypothesis using a logit model. Results are presented in Table 14. Model 1 includes only the key ethnic fractionalization variable. The ethnic fractionalization measure is both strongly significant and negatively correlated with expressed preference for co-ethnics. With the ethnic fractionalization measure set at the mean, there is a 53.7% chance a respondent will prefer a co-ethnic as president. However, if we increase ethnic fractionalization by one standard deviation (.448 to 0.741) then the probability of the respondent preferring a co-ethnic drops to 45.4%. Thus increasing ethnic fractionalization by one standard deviation decreases the probability a respondent will prefer a co-ethnic by 8.3%. Table 14 – Public preferences for co-ethnic leadership Preferences for Co-Ethnic Candidates (logit) Model 1 Model 1 Estimate (std. err) Estimate (std. err) Ethnic Fractionalization -1.14*** (.13) -1.31*** (.16) Male .02 (.08) Age -.10* (.05) Education -.12*** (.04) Islam .34*** (.13) Javanese -.35*** (.09) Urban -.43*** (.09) Constant .66*** (.07) 1.47*** (.24) Observations 2994 2873 R2 0.0198 0.0437 This is a substantial impact, but the simple result is not altogether surprising. Many of the homogenous districts in the sample are predominantly Javanese, which is also the ethnic group most likely to be represented in the executive. Model 2 adds the control variables. Again, the ethnic fractionalization measure is strongly significant and negatively correlated with expressed preference for co-ethnics. With all the controls set at the mean value, increasing ethnic fractionalization by one standard deviation changes the 175 probability of co-ethnic preference from 53.8% to 44.3%. This decrease of 9.5% is, again, substantial. The effect of the control variables is mixed. Neither age nor gender has any significant impact on the dependent variable. Both education and urbanization have the expected negative effect on preferences for co-ethnics. A one standard deviation from mean in education lowers the probability of a co-ethnic preference by 3.4%, while an urban voter is 10.5% less likely to prefer a co-ethnic than a rural counterpart. The two surprising results involve the effect of Islam and Javanese identity. A Muslim is 8.5% more likely to prefer a co-ethnic president than a non-Muslim, while an ethnic Javanese is 8.7% less likely to prefer a co-ethnic president than a non-Javanese. The latter is both the most relevant and the most confounding. It suggests that the voters with the strongest ethnic preferences for president are the least likely to ever see a co- ethnic win the position. This does dovetail with previous literature suggesting Indonesia’s ethnically dominant Javanese less cognizant of ethnic politics than other groups (Emmerson 1976). It also adds confidence that the measure captures preferences for co- ethnics politicians in general, rather than preferences for simply co-ethnic presidential candidates. The relationship between ethnic dynamics and co-ethnic preferences become clearly visible when we examine provincial aggregates. Provincial ethnic fractionalization is plotted on the X-axis, aggregate percentage of co-ethnic preferences on the Y-axis. While the provincial fractionalization measure is not as powerful a predictive measure as the district fractionalization, I use provincial aggregates here to simplify data presentation. Figure 25 presents results. The provinces with the highest percentages of voters expressing preferences for co-ethnics are West Sumatra and South Kalimantan. These provinces are both relatively homogenous and non-Javanese. The homogenous Javanese provinces of Central Java and Yogyakarta also have high percentages preferring co- ethnics. In the lower right corner sit the diverse provinces that report low levels of ethnic preferences, including South Sulawesi, East Nusa Tengerra, South Sumatra, and, Jakarta. While the ethnic fractionalization measure does not account for all variation in ethnic preferences, there is a clear trend. 176 Figure 25 – Public preferences for co-ethnic leadership across provinces Ethnic diversity does not generate a hardening of ethnic biases. In fact, the opposite appears to be true: ethnic diversity produces a moderation of ethnic biases as they relate to the political sphere. Voters in diverse districts are less likely to express a preference for co-ethnic candidates. This does not mean they are necessarily voting across ethnic lines during legislative elections. It does raise serious doubt that a simple communal voting story can account for the dynamics we see across Indonesian districts.207 Communal voting revisited Ethnic diversity does not appear to produce a clear ‘ethnification’ of political campaigns, does not produce a political preference for co-ethnic candidates, and can account for only minimal variation in preference voting. Still, the communal voting model cannot be 207 Though the strength of the correlation is striking, the relationship between ethnic diversity and political tolerance is not without precedent in the region. In neighboring Malaysia, the multi-ethnic Barisan Nasional tends to turn in its strongest electoral showings in those districts with mixed populations while suffering losses to more stridently ethnic parties in homogenous districts (G. K. Brown 2005). North Sumatra West Sumatra South Sumatra Lampung Banten Jakarta West Java Central Java East Java East Nusa Tengerra South Kalimantan South Sulawesi Jogyakarta 30 40 50 60 70 % Pre fer Co -Et hn ic P res ide nt 0 .2 .4 .6 .8 1Ethnic Fractionalization Co-Ethnic Preference by Ethnic Fractionalization 177 entirely dismissed. A politician’s anticipated access to rents makes a voter more likely to attach themselves to a candidate, but it does not dictate what type of candidate a voter will attach herself to. Voters in diverse districts are less likely to hold strong preferences for co-ethnics; however, they may still be voting for co-ethnics. Rather than diversity directly producing co-ethnic preferences, voters may simply vote for co-ethnics for reasons of familiarity and exposure. The positive correlation between local candidates and preference voting suggests voters prefer familiar candidates. Candidates, likewise, prefer to leverage kinship and religious networks that tend to follow ethnic lines, increasing the exposure of voters to co-ethnic candidates. The story of a North Sumatra PD candidate illuminates this attenuated communal dynamic.208 The candidate in question was a prominent provincial official in the party from a Malay background. She had run for national office in 2004. It was stressed to me that the party did not pick candidates on an ethnic bases or generally strategize along ethnic lines. However, when asked what she found challenging about campaigning, she responded that the language barriers inhibited her campaign efforts. Candidates, she explained, could more easily connect with voters when they shared the local dialect. Even if a candidate had been living in Jakarta for a long period – which many national candidates had – knowledge of ethnic dialects and customs eased campaign efforts. The more subtle communal voting mechanism has validity. Exposure and familiarity undoubtedly matter for candidates. However, exposure and familiarity do not directly determine the importance of personality politics. Rather, rent opportunities are the major factor focusing voters’ attention to personal appeals, both because it makes promises of future assistance credible and draws higher numbers of competitors into the electoral arena. Communal dynamics may help determine which candidate a voter attaches herself to, but rent opportunities determine why voters bother making that choice in the first place. Conclusion This chapter finds that, in the Indonesian context, rent opportunities lead to a personalization of political dynamics. Where rent opportunities are high, candidates can 208 Interview with PD DPRD candidate, Medan, (4 March 2009) 178 credibly promise to reward (and punish) voters following the election. Voters in high rent areas have an incentive to attach themselves to a patron. Higher rates of preference voting across electoral districts reflect these exchange relationships. This finding fits nicely into my causal argument linking rent opportunities to electoral fragmentation. Personalization localizes politics, just as localization helps to personalize politics. Robust exchange relationships between candidates and voters diminish the saliency of programmatic and ideological appeals. While all parties adjust themselves to these local dynamics, it is the larger parties that are most detrimentally affected. When the large parties cannot rely on their national brand, small parties are able to compete on a more equal footing. The secondary effect of the exchange relationships is the growth of minor parties and the fragmentation of local political systems. 179 Chapter 7 – Party System Size Indonesia’s efforts to engineer party system outcomes have been the most extensive in the Asia Pacific (Reilly 20f06, 132). A variety of tools have been employed to achieve a few core goals. First, institutional engineers have tried to control the number of partisan actors by manipulating legislative thresholds, district magnitudes, and registration rules. Second, regional registration requirements have been written so as to block the emergence of regional and ethnic parties. Though the recent reforms have been enacted by self-interested incumbents that see advantage in tightening electoral laws, they have been cheered on by outside actors in civil society and the media. The venture in institutional design has had mixed results. The effective number of legislative parties has gone down, though the effective number of electoral parties has continued to expand. Indonesia now finds itself the unenviable position of having a fragmented national legislature and a surprisingly disproportional electoral system. There are also signs that the law requiring parties to organize across the archipelago is not producing the desired ‘nationalization’ of political competition. Voters may be forced to select national parties, but a review of recent results suggests ethnic diversity also plays a role in electoral politics. The districts with a large number of ethnic groups also have the most fragmented party systems. For instance, in the most recent election district-level party system size ranged from a high of 13.6 in ethnically heterogeneous East Nusa Tenggara I to a low of 4.3 in ethnically homogenous Bali. Holding several key variables constant, I find that moving from the most homogenous to the most heterogeneous districts produces an increase of 3.1 parties in 2004 and an increase of 1.9 parties in 2009.209 This could indicate national parties are merely a façade masking potentially de- stabilizing ethnic competition. Why do diverse districts produce fragmented party systems? At first glance, outcomes seem to conform with an established literature connecting ethnic diversity to fragmented party systems. Yet none of the assumed mechanisms seem to apply in Indonesia: parties and candidates are cautious in their public use of the ethnic issue, 209 See Appendix G, Section 1. 180 public attitudes in diverse areas are surprisingly tolerant, and there is no evidence that ethnic groups vote as a coherent bloc. There are, in short, few signs of ethnic politics. I argue that the correlation between ethnic diversity and the number of parties is caused indirectly through the mechanism of rent opportunities. The opportunity to abuse state resources for personal and/or political ends alters the behaviour of candidates and voters. Where rent opportunities are high, locally oriented candidates running under minor party labels are able to attract voters with a clientelistic message. As a result, voters disperse their votes across a wide range of major and minor parties, fragmenting the local party system. The chapter proceeds as follows. First I review the literature on party systems, ethnic diversity, and ethnic voting. Second, I lay out the rent opportunities theory of party system fragmentation. Third, I explain how the competing theories propose different mechanisms with distinct observable implications that can be investigated empirically. The fourth section tests hypotheses, with a focus on intra-district dynamics. Finally, I conclude with a summation of findings and a discussion of significance. Ethnic diversity and party systems: existing explanations The curious treatment of ethnic parties How does ethnic diversity affect party system size when ethnic parties have been effectively prohibited? The first place to look is the large literature that connects ethnic diversity and party system size (Brambor, W. R. Clark, and Golder 2005, 2007; W. R. Clark and Golder 2006; Cox and Amorim Neto 1997; Cox 1997; Filippov, Ordeshook, and Shvetsova 1999; Geys 2006; Mozaffar, Scarritt, and Galaich 2003; Ordeshook and Shvetsova 1994; Stoll 2007; Vatter 2003). In short, ethnically diverse countries produce fragmented party systems. The relationship between ethnic diversity and party system size is mediated by electoral institutions: the effect of diversity is large in ‘permissible’ systems (e.g. proportional with high district magnitudes) and small to in ‘restrictive’ (e.g. single-member district plurality) systems. Party system size, then, is the product of the interaction of social heterogeneity (as measured by ethnic diversity) and electoral institutions. 181 With the exception of Stoll and Mozaffar at al., the issue of how and why ethnic identities affect voter choices is left unexplored and thus the role of ethnic parties remains unclear (Mozaffar, Scarritt, and Galaich 2003; Stoll 2007). The canonical works are curiously vague on the subject. Ordershook and Shvestova, the pioneers of the interactive approach, preface their discussion of ethnic heterogeneity with the comment “We need not review the innumerable essays that document the influence of ethnicity on politics” (Ordeshook and Shvetsova 1994, 107). Later they note their measurement of diversity is meant to capture the number of possible groups within an electorate (Ordeshook and Shvetsova 1994, 108-9). But the authors remain reserved on whether specific ethnic groups lead to ethnic parties: [Our] purpose here is not to ascertain precisely how ethnic heterogeneity influences party systems. Rather, we merely want to determine whether the influence of a single institutional variable, district magnitude, on the number of parties is better described if we take a simple characterization of a society’s ethnic structure into account, with the understanding that there is considerable room for additional refinements in the conceptualization and measurement of variables. (Ordeshook and Shvetsova 1994, 110) Part of Ordershook and Shvestova’s mechanism suggests a straight-forward argument in which more groups potentially lead to more parties based around group lines, yet they stop short of making this connection.210 The cryptic treatment of ethnic parties is again present in Amorim Neto and Cox (1997). The authors use the effective number of ethnic groups as a measure of social cleavages, which they define as “enduring social differences that might become politicized or might not” (Cox and Amorim Neto 1997, 152). Cleavages can translate into partisan preferences which, depending on contextual circumstances, may translate into votes. Social cleavages thus determine the ‘need’ for parties. The causal mechanism assumes distinct parties form around the social cleavages; however, the authors are careful not to suggest that ethnic diversity leads to ethnic parties: [We] view the effective number of ethnic groups as a crude proxy for social diversity lato sensu. Thus, we do not necessarily expect that more ethnic groups lead to more ethnically-based parties. To begin to get predictions of this kind, one 210 In a later collaboration the authors they begin to fill conceptual gap. The authors state that, in high district magnitude districts, “parties that cater to specific minorities have greater incentive to form and less incentive to coalesce with other parties” (Filippov, Ordeshook, and Shvetsova 1999, 13). 182 would have to take on the issue of cross-cutting cleavages in some fashion - which seems a hard task. (Cox and Amorim Neto 1997, 166). Thus Amorim Neto and Cox get around the problem of explaining how and why voters select co-ethnics by simply suggesting ethnic voting is not a necessary part of their causal mechanism. Clark and Golder’s discussion of social cleavages and party system outcomes is heavily influenced by Duverger. Social cleavages produce ‘spiritual families’ that can be particicized (W. R. Clark and Golder 2006, 681). Social cleavages also “represent ‘natural constituencies’ that generate particular policy demands” (W. R. Clark and Golder 2006, 682). While parties form around lines of cleavage, it is unclear if the parties will be ethnic. The authors state, “Although the number of ethnic groups represents just one element of social heterogeneity, it is a proxy that all previous analyses use and, therefore, provides the best means for comparing our results with existing findings” (W. R. Clark and Golder 2006, 696). Like those before them, Clark and Golder are vague as to whether or not ethnic diversity leads to ethnic parties, though it is clearly implied in their discussion. Mozaffar et al. tackle the issue directly by shifting the focus from ethnic diversity to group concentration (Mozaffar, Scarritt, and Galaich 2003). As they point out, previous literature simply assumes that each ethnic group is large and cohesive enough to support a party on its own (Mozaffar, Scarritt, and Galaich 2003, 380-1). They suggest diversity should have a reductive effect on party system size as most small groups cannot support a party on their own. Group concentration lowers the cost of mobilizing around ethnicity and thus facilitates the particization of group (Mozaffar, Scarritt, and Galaich 2003, 382-3). Brambor et al. challenge the finding that diversity reduces party system size, demonstrating the Mozaffar et al. conclusion was based on faulty statistical procedures (Brambor, W. R. Clark, and Golder 2007). In the process, however, they implicitly endorse the ethnic party mechanism. The party systems literature leaves us in a strange position. Existing theory clearly implies a three part mechanism: 1) ethnic diversity is a dominant a social cleavage; 2) parties form to represent voters along this cleavage; 3) mobilization of these groups fragments the party system. The theory suggests mobilization occurs around ethnic 183 parties; nonetheless, authors have tended to either ignore the subject or suggest some other undefined mechanism is at play that negates the importance of ethnic voting. As such, we need to go beyond the party systems literature to determine the conditions under which diversity might be expected to produce fragmented party systems. Ethnic identities and voting How electoral choices are shaped by a voter’s ethnic background remains a matter of debate. Differences in the accounts of ethnic voting have consequences for our expectations of how diversity affects party system outcomes, especially in countries that do not have ethnic party systems. I review three existing answers that emphasize the role of policy preferences, group esteem, and cognitive short-cuts.211 First, a preferences story proposes that each ethnic community socializes their members in such a way as to generate similar views on policy questions (Alesina, Baqir, and Easterly 1999; Rabushka and Shepsle 1972). These views tend to diverge from people outside the group. The policies on which ethnic groups diverge could be distinct from a specific ‘ethnic’ concern. For instance, a member of one ethnic group may hold policy views on wealth redistribution that closely match co-ethnics but diverge from positions held by members of other ethnic communities. How the homogenization of preferences within an ethnic community occurs remains under-explored. It is possible that the prevalence of intra-group communication facilitates a convergence of views within a group. Or there may be different values imparted through distinct educational, religious, or associational practices. No matter the sociological mechanism at work, the basic assumption remains that the experience of living within an ethnic community instils policy preferences that affect electoral choices. A second account suggests voters derive psychological benefit from voting for co- ethnics. Horowitz explicitly challenges the notion that the appearance of ‘ethnic’ political claims simply represents the clustering of policy preferences (Horowitz 1985, 345). In divided societies, voters view elections less as an expression of policy preferences and more as an act of group allegiance. Horowitz’s language evokes a spiritual battle: the ethnic cause has a defined “ethnic enemy,” carries an “element of sacredness,” and a 211 This division of the literature is borrowed from Ferree (Ferree 2006). 184 voter that fails to support their co-ethnics “may carry an indelible stigma” (Horowitz 1985, 344). Rather than evaluating policy platforms, then, voters often “choose, in effect, not to choose” and instead lend their support to a clearly ‘ethnic’ political option (Horowitz 1985, 323). A third explanation of ethnic voting highlights the role of cognitive short-cuts. Chandra describes how ethnic identities are useful to voters and elites facing information constraints (Chandra 2004, 2007). Voters’ desire access to state favours and elites want reliable voting blocs. Because ethnic traits are visible and widely recognized, both voters and elites use ethnic identity as a cue to determine who can be counted on to deliver either favours or votes. In other words, voters expect co-ethnic politicians to deliver favours, and thus exchange votes for a candidate’s promise of support once in office. Ferree expands the logic beyond the retail exchange of favours for votes (Ferree 2006). She posits that voters use readily available information to develop ‘ethnic party profiles’ that allow them to determine whether or not a party can be relied upon to advance the interests of a given ethnic group. For both Ferree and Chandra, ‘ethnic voting’ occurs because voters use ethnic cues as a cognitive short-cut when evaluating political options. Party systems and diversity: mechanisms reconsidered The different accounts of ethnic voting imply different conditions under which diversity fragments the party system. In the preferences story, ethnic parties are not a necessary piece of the causal mechanism connecting diversity and party system outcomes. A multiplicity of groups within a district should fragment the party system if the parties hold distinct policy positions and the lines of distinction match up with ethnically derived clusters of policy preferences. For example, if there are three parties spaced equidistant along a left-right axis, and three ethnic groups with redistribution preferences that cluster around the same dimension, ethnic block voting is a plausible outcome. Depending upon the configuration of electoral institutions and the size of each group, block voting could fragment the party system, even the absence of distinctly ‘ethnic’ parties. The social psychological story clearly implies that diversity will produce party system fragmentation if ethnic parties exist. Where one party claims to represent each ethnic group, the logical conclusion is a ‘census election’ in which electoral results closely follow demography. Even if groups split their vote between distinct ethnic 185 options, the existence of multiple groups within a district could fragment the electoral vote. Yet it is unclear what outcome would occur in the absence of ethnic parties. The argument implies a heavy emotional commitment in the victory of the party, which might still occur in non-ethnic or multi-ethnic party systems. Nonetheless, the approach provides little guidance on how blocks may spread themselves out across a system lacking ethnic parties and thus offers no clear prediction about the relationship between diversity and party system size. The cognitive approach yields predictions similar to the social psychological approach. When clearly ethnic options exist, district level diversity should fragment the party system. It is even possible that ‘non-ethnic’ parties at the national level take on an ‘ethnic’ component sub-nationally due to the preponderance of regional leadership control by a particular group. Where parties are non-ethnic at both levels, there is little clue as to how diversity can affect party system outcomes. Even if voters are ethnically minded and keep party profile as described by Ferree, the existence of group voting does not tell us anything regarding how groups will be dispersed across parties (Ferree 2006). The Indonesian context The conditions under which district-level diversity produces a high number of political parties are onerous. A survey of the Indonesian context suggests none of the three potential stories fit. Even if preferences within ethnic groups do converge, it is difficult to say how they would map on to the cotemporary party system. The line of conflict with the most coherence is the secular-religious axis, with PDI-P on the secular end of the spectrum and PKS on the religious. Ethnic groups could have distinct preferences on the role of Islam in public life. There are some obvious problems with the story however. In 2009, for instance, the two largest players (Golkar and PD) were ideologically vague parties that promoted themselves nationally either on the basis of competence (Golkar) or leadership (PD). Since 1999 the party system has been moving toward ‘centrism’ on religion (Tanuwidjaja 2010). Supposedly secular parties have become more open to ‘Muslim’ issues while religious parties have softened their image on controversial issues. If there was going to be an election where preferences led to clear patterns of party support it would have been 1999, when policy positioning along religious and reformist axes approached coherence. However, no district-level correlation between ethnic 186 diversity and party system fragmentation existed in that year. In short, the preferences story simply does not appear plausible in the Indonesian setting. The cognitive and social psychological stories also run into problems. It is possible that district-level ethnic divisions animate party competition even in the absence of ethnic parties. Indonesia itself has a history of ethnic battles for sub-national power in which local party leadership is dominated by one group (Liddle 1970). However, the mere existence of ethnic politics in diverse districts does not imply diverse districts will have more parties than homogenous districts. For diversity to produce fragmentation we must also assume ethnically homogenous areas are shaped by a cleavage structure that simplifies party competition. Even in 1950s Indonesia, this was not the case. Homogenous Javanese provinces, for instance, were divided by class and religion and produced party systems even more fragmented than many diverse areas.212 In sum: existing theories do not lead to clear predictions about the relationship between ethnic diversity and party system size in Indonesia. All three stories assume that members of an ethnic group vote as a block for the same party or parties, an assumption I henceforth refer to as the Communal voting model. Though communal voting is possible explanation for Indonesian outcomes, I have explained why this is unlikely in this particular case. To understand why the correlation between ethnic diversity and party system fragmentation exists we must look beyond the constraints of the Communal voting model. Rent opportunities, ethnic diversity, and party system size To understand why Indonesian voters support a wide array of parties in some areas it is essential to understand what factors motivate political competition. The central claim made in this dissertation is that the opportunities to manipulate sub-national resources serve as a key motivation for much political organization in Indonesia. For reasons independent of the electoral process, diverse regions in Indonesia tend to have high rent opportunities. These rent opportunities directly shape the goals and strategies of both 212 In 1955, the average effective number of electoral parties per district was approximately 3.95. The average ENEP in the three Javanese dominated districts on Java (Yogyakarta, Central Java, and East Java) was 4.42. Data from Kevin Evans (Evans 2003). Note: Evans’ 1955 data is aggregated by 1999 provincial boundaries and thus does not match up to the 1955 electoral districts in a few cases. 187 elites and voters. First, there are higher rates of candidate entry in high rent areas. The state dominates economic life and attracts the attention of ambitious elites. Consequently, a high number of elites enter the political sphere looking to access local rents. Second, viable elites affiliate with minor parties. As attention shifts from national issues to local rent distribution, the political programmes of the major national parties carry less appeal to voters and elites. Elites feel free to use party labels opportunistically, seeking out labels that minimize the costs associated with affiliation. Third, voters anticipate local rent sharing and support parties based on their belief in the party’s candidate to deliver particularistic goods. Focus on candidates rather than party platform leads voters to support an array minor party labels. All three factors – increased entry, opportunistic affiliation, and candidate-centered competition – combine to fragment the local party system. My argument linking ethnic diversity and electoral outcomes assume neither ethnic parties nor communal voting. Voters may in fact vote for co-ethnics in diverse areas. Still, co-ethnic voting alone cannot explain the party system outcomes we see. Instead, I posit that the broader economic context is the key regional factor driving elite and voter strategies. Hypotheses The discussion above contains two theoretical stories linking ethnic diversity and party system fragmentation. The rent opportunities model links party system fragmentation to ethnic diversity indirectly through the mechanism of rent opportunities. This model leads to Hypothesis 1: H1: The number of parties is positively associated with rent opportunities In contrast, the communal voting model posits a straight-forward relationship between ethnic group membership and partisan support. H2: The number of parties is positively associated with ethnic diversity Given that both causal stories expect a correlation between ethnic diversity and party system size, a straight-forward investigation of party system size and ethnic diversity at 188 the level of the electoral district cannot tell us much about the validity of the two competing theories. In order to test the validity of each theory I parse out additional observable implications. First, if the rent opportunities model is correct rents should have an independent effect on party system size even in the absence of ethnic diversity. H3: The number of parties is positively associated with rent opportunities even if ethnic diversity is low The correlation between rent opportunities and ethnic diversity is close but not exact. Some homogenous national-level electoral districts have more rent opportunities than others. Furthermore, the rent opportunities model makes specific predictions about intra- district dynamics. The two theories lead to distinct hypotheses regarding electoral politics within an electoral district: in the communal voting story, party system size is kept in check within the ethnic group, but the presence of multiple groups ends up fragmenting the vote. The rent opportunities theory sees party systems fragmenting because economic conditions generate clientelistic behaviour even within the context of intra-ethnic group competition. Thus the rent opportunities theory predicts a relationship between rents and party system size even within homogenous municipalities. A complementary hypothesis considers the effect of ethnic diversity in the absence of rent opportunities: H4: The number of parties is positively associated with ethnic diversity even if rent opportunities are low The communal voting story assumes each group has distinct preferences (or simply a preference for co-ethnics) which leads them to vote for different political candidates or parties. Because the assumptions underlying the model do no rely on access to rents, we might expect to see a relationship between ethnic diversity and electoral fragmentation in the absence of rent opportunities. More specifically, we would expect ethnic diversity to correlate with electoral fragmentation even in Indonesia’s non-partisan upper house 189 elections, which are largely detached from the competition for sub-national resources. The proceeding empirical section sets up a series of statistical tests to investigate the validity of the competing interpretations of electoral system fragmentation. Theory testing: district level party systems Hypotheses 1 and 2: effective number of electoral parties This section offers an initial test of Hypotheses 1 and 2 through an examination of national electoral districts. The key dependent variable is the effective number of electoral parties in the national district (ENEP-ND]. To test Hypothesis 1 I use percentage of state employees in the modern sector as the key proxy of rent opportunities. The statistical model controls for district magnitude and urbanization. Since elections in Aceh in 2009 took place under distinct laws allowing local parties, I also add an Aceh dummy for the most recent election. Table 15 presents results.213 In 2009, coefficients for two of the control variables – urbanization and Aceh – are negative and strongly significant. Logged district magnitude has no effect. The civil service size variable is positive and does reach standard levels of statistical significance.214 An increase of one standard deviation in civil service size increases ENEP-ND by 0.48. Using Clarify, I generate predicted values for electoral districts around the 25% and 75% percentile in terms of civil service size. With all variables set to the mean and Aceh set to 0, the model predicts that an electoral district in East Java (4.7% civil servants) will have 8.3 electoral parties. In contrast, a district in South Sulawesi (13.7% civil servants) will have 9.0 electoral parties.215 The size of the effect is modest but hardly trivial.216 213 For reasons of space I include only results from 2009. Results from 2004 are available in Appendix G, Section 2. 214 The result holds for alternative specifications of rent opportunities. See Appendix G, Section 3. 215 The predictions closely match the empirical outcome in 2009: the average district in East Java had 8.4 electoral parties while the average district in South Sulawesi had 9.0. 216 Running a similar test reveals that the magnitude of the effect was larger in 2004: it predicts an electoral district in South Sulawesi would have about 1.5 more electoral parties than an electoral district in East Java. 190 Table 15 – Determinants of party system size Determinants of Party System Size, 2009 – National Legislature (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 – Hypothesis 1 Model 1 – Hypothesis 2 Model 3 – Full Model Civil Service Size 0.08** (0.03) -0.003 (0.04) Ethnic Fractionalization 2.03*** (0.51) 2.06*** (0.67) District Magnitude -0.22 (0.58) -0.15 (0.53) -0.16 (0.55) Urbanization -3.38*** (0.78) -4.18*** (0.70) -4.21*** (0.78) Aceh -4.65*** (1.11) -4.51*** (1.02) -4.50*** (1.05) Const 9.70*** (1.34) 9.68*** (1.11) 9.72*** (1.27) Observations 77 77 77 R2 0.4044 0.4747 0.4747 It is noteworthy that electoral fragmentation is a phenomenon strongly driven by minor party voting.217 Though not inconsistent with a communal voting story, the correlation between minor party voting and electoral fragmentation is an essential aspect of the rent opportunities model. According to the logic of the rent opportunities story, electorally viable elites avoid major parties because their goals are local rather than national, major party affiliation is costly, and major party labels offer few electoral benefits when voters are oriented toward accessing local rents. Evidence from both 2004 and 2009 is consistent with the story. The surge in minor party support does not completely displace the major parties; rather, minor party success adds more viable partisan components to the system.218 217 The designation of ‘major party’ status varies between elections. The criteria for inclusion are: 1) over 5% of the national vote; 2) a recognized, viable presidential candidate. Note: the criteria used here are looser than those employed in Chapter 4, which limited ‘major party’ status to only those parties that had achieved 5% of the vote in the previous election. See Appendix G, Section 4. 218 The one exception is Bangka Belitung in 2004, where the success of PBB (21%) had a consolidating effect on the party system (ENEP: 7.5). In three other cases that year high minor party voting correlated with a below average party system size (NTT I, NTTII, Central Sulawesi), though in all three the consolidation of the system occurred due to the lingering electoral strength of Partai Golkar rather than a strong breakthrough for any particular minor party. 191 To test Hypothesis 2, Model 2 replaces the rent opportunities proxy with an ethnic fractionalization variable. The two key control variables remain statistically significant, with minor changes in the size of the correlation coefficients. The ethnic fractionalization variable is positively signed and strongly significant. I use Clarify to generate predicted values around the 25% and 75% percentile on the ethnic fractionalization measure. The model predicts a relatively homogenous district like Central Java 8 (0.12) to have 8.0 electoral parties, while a diverse district like Jambi (0.78) to have 9.3 electoral parties. Notably, examining the differences in R2 values reveals that Model 2 explains more variation in the dependent variable than Model 1. In Model 3 I add both the ethnic fractionalization and civil service size variables. The coefficient of civil service size changes signs and shrinks below statistical significance. The ethnic fractionalization variable, however, stays positive and remains statistically significant.219 In this simple test there is no evidence the rents variable has an independent effect. Evidence from this section includes support for both Hypotheses 1 and 2. Civil service size does correlate with the number of national electoral districts in 2004 and 2009. The same can be said of ethnic diversity. When both are added to the model, only ethnic diversity remains statistical significance. This result, however, could be a product of measurement error. Civil service size captures only the resources element of the rent opportunities concept; it is simply assumed constraints on elite behaviour are uniform, though we know from Chapter 3 that constraints do in fact vary with ethnic diversity. This initial round of testing demonstrates that both H1 and H2 are plausible but does not provide a decisive test of the two theories. Hypothesis 3: the effect of rent opportunities in the absence of ethnic diversity Only limited statistical testing of the competing theories can be accomplished using national electoral districts as the primary dependent variable. Given the multicolinearity of the key independent variables and the small sample size (69 observations in 2004; 77 in 2009), it difficult to sort out the independent effects of rents and ethnic diversity on party system size. Disentangling the effect of each variable requires digging into sub- 219 A similar result occurs in 2004. See Appendix G, Section 2. 192 district dynamics. National electoral districts in Indonesia are typically made up of multiple municipal administrative units (kabupaten/kota). While a national electoral district has one ethnic fractionalization score, the municipal units that constitute the district can vary in their ethnic structures. Looking at sub-district voting patterns allows us to isolate the various mechanisms of electoral fragmentation. I use sub-national election results from the 2004 election collected by the Indonesian Community for Democracy (Komunitas Indonesia untuk Demokrasi, KID). The dataset contains results from every electoral district in all of Indonesia’s municipal governing bodies. Because Indonesian voters tend to support the legislative party at all three levels of governance (municipal, provincial, and national), municipal results provide a close approximation of national-level behaviour.220 There were 69 national electoral districts in 2004. Two of these were in the Special Capital Region of Jakarta and thus did not have municipal elections.221 The remaining 67 contain a total of 433 municipalities. Each municipality is sub-divided into electoral districts, with the number of electoral districts per municipality ranges from 2 to 7. For each observation I calculate the party system size in the municipal district (ENEP- MD), which I use as the key dependent variable. In total there are 1744 municipal districts. The average ENEP-MD is 7.0 with a standard deviation of 2.5. To test the communal voting theory I add municipal-level ethnic fractionalization. There are two noteworthy problems with the measure. First, the data I collected measured ethnic structure at the municipal level but my dependent variable measured sub- municipal electoral outcomes. Due to the fact that ethnic groups tend to be clustered within a municipality, the fractionalization score I assign to each unit is likely greater than the ethnic diversity within the municipal electoral district itself. Second, Indonesia experienced considerable change to its municipal boundaries between 2000 (the year of the census) and 2004 (the year of the election). Many municipalities were split in a process known as ‘pemekaran’ (blossoming). Where I lack data on the ethnic structure of the new municipality I assign a fractionalization score from the originating municipality. 220 For further evidence of straight ticket voting, see Appendix G, Section 5. 221 Jakarta is technically a province. The administrative units of the province did not have separate elections in 2004. 193 Both problems with the data cause an important overestimation of the ethnic fractionalization values. As a proxy for rent opportunities I use civil service size. A district magnitude variable captures the number of seats in the electoral district. The district magnitude variable is logged because increases in the independent variable should have a larger impact on party system size when district magnitude is low that when it is high. An additional variable tests for an interactive relationship between ethnic diversity and district magnitude. Lastly, in order to capture urbanization affects, I add a dummy for all municipalities categorized as ‘kota.’ Table 16 contains results. Due to data constraints I drop all observations in which the largest ethnic group was “other.”222 Model 1 presents results from a simple four variable test with no interactive term. The kota control is positive and strongly significant. This result captures the partial breakthrough of two insurgent parties – PD and PKS – among the urban middle class. All else being equal, an urban district has about 0.79 more electoral parties than a rural district. The district magnitude variable, however, was not significant. 222 Most often the “other” category is simply a collection of smaller groups. In a few municipalities, however, it is clear that “other” is in fact one large group. The exclusion results in qualitative changes in coefficient sizes but does not affect the statistical significance of any results displayed. For more, see Appendix G, Section 6. 194 Table 16 – Municipal level party system size Determinants of Party System Size – Municipal Legislature, 2004 (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 – Simple Model Model 2 – Interactive Model Model 3 – Homogenous Sample Civil Service Size 0.07*** (0.01) 0.07*** (0.01) 0.09*** (0.02) Ethnic Fractionalization 2.57*** (0.22) 1.13 (1.55) District Magnitude 0.18 (0.21) -0.13 (0.39) -0.20 (0.33) Ethnic X Magnitude 0.70 (0.74) Kota 0.70*** (0.15) 0.70*** (0.15) 1.14*** (0.24) const 4.78*** (0.45) 5.42*** (0.83) 5.39*** (0.72) Observation 1550 1550 630 R2 0.1860 0.1865 0.0778 Both of the key independent variables are positively signed and strongly significant. A one standard deviation increase in civil service size produces an increase of 0.5 electoral parties while a one standard deviation increase in ethnic fractionalization increases the number of electoral parties by 1.0. Alternative measures of the rent opportunities concept yield similar results with larger predicted effects.223 For instance, moving from the minimum to maximum number of candidates-per-seat produces 1.7 more parties. To facilitate interpretation I generate predicted values in four municipalities. Results appear in Table 17. Each municipality was selected because it had ethnic fractionalization and civil service size values at or very near to either the 25th or 75th percentiles. In the Low Fractionalization / Low Rents case of Sampang, the model predicts an ENEP-MD of 5.7. In contrast, the High Fractionalization / Low Rents municipality of Tangerang is predicted to have 7.3 electoral parties. Moving to the Low Fractionalization / High Rents municipality of Sinjai, the model predics the existence of 223 Additional measures include the number of national candidates-per-seat and the provincial fractionalization measure. See Appendix G, Section 7. 195 6.3 parties. Lastly, in the High Fractionalization / High Rents municipality of Tanjung Jambi is predicted to have 8.2 electoral parties. The magnitude of the rent opportunities variable is sizable but noticeably smaller than that of the ethnic diversity variable. For instance, the difference between Tanjung Jambi and Sinjai is 1.9 while the difference between Tanjung Jambi and Tangerang is only 0.9. The initial simulation, then, suggests municipal fractionalization is the more important variable. Nonetheless, the simulations also indicate that the municipalities with the highest electoral fragmentation are those with a combination of both ethnic diversity and rent opportunities. Table 17 – Predicted values of municipal party systems Predicted Values in Four Municipalities Province Municipality Ethnic Fractionalization Civil Service Size (% non- agricultural) Predicted Party System Size East Java Sampang 0.09 5.0 5.7 Banten Tangerang 0.75 3.2 7.3 South Sulawesi Sinjai 0.08 13.2 6.3 Jambi Tanjung Jambi 0.76 13.4 8.2 Model 2 adds the interactive term, thereby providing a more precise test of the established theory of ethnic voting. It is difficult to interpret the municipal fractionalization variable due to the inclusion of the interactive term. Using Brambor et al.’s method, Figure 26 plots the marginal effect of a one unit change in the ethnic diversity variable dependent upon district magnitude (Brambor, W. R. Clark, and Golder 2005).224 A weak relationship is present: ethnic diversity has a greater impact on electoral fragmentation when district magnitude is relatively high. With a district magnitude of 5, a one unit change in the ethnic diversity variable produces about 0.2 more parties. A similar change at a district magnitude of 12 produces about 0.3 more parties. This finding conforms to established literature and runs counter to the rent opportunity model. 224 For this test I re-ran model 3 with minor modifications. I omitted observations where the largest group was ‘other.’ Also, for reasons of presentation I used the effective number of ethnic groups rather than ethnic fractionalization. This transformation allows for an easier interpretation of a ‘one unit change’ and does not substantially alter results of model. See Appendix G, Section 6. 196 Figure 26 – Marginal effect of ethnic diversity on party system size The previous tests are hampered by two problems. First, the municipal fractionalization measure overestimates the ethnic diversity within a municipal electoral district. Second, the tests suffer from a measure of multicollinearity. Municipal fractionalization correlates with all of the rent opportunity variables, thus results for each of the individual variables need to be treated with a degree of caution. To ensure that an independent impact of the rent variables exists I limit the sample to only homogenous municipalities, defined as those that have a group with 90% or more of the population. In this context we would not expect municipal fractionalization to have any impact, so I am able to drop the variable from the model. Results appear in Model 3. Again, the rent opportunities variables stay positive and strongly significant. A one standard deviation increase in the civil size variable results in an increase of 0.62 electoral parties. Similarly, the model predicts that moving from Sampang’s civil service size to Sinjai’s increases electoral fragmentation by 0.75. Civil service size strongly correlates electoral fragmentation at the sub-district level. Model 3 provided a decisive test of Hypothesis 3. I find that that the key rent opportunities proxy correlates with party system fragmentation even in the absence of 197 ethnic diversity. Given that the municipal fractionalization variable substantially overestimates the level of ethnic diversity within a municipal-level electoral district it is difficult to measure the direct effect of diversity. There are signs of a direct effect and perhaps even an interactive relationship with district magnitude. Even if a direct relationship exists, however, it would not negate strong support for Hypothesis 3. To sum up the section: while indicators of rent opportunities impact municipal party system size, so does the municipal level ethnic diversity. As it stands, the result is puzzling; there is little evidence suggesting block voting by ethnic groups, which is the mechanism we expect to link ethnic diversity and party system size. It is likely that the municipal fractionalization variable substantially overestimates the level of ethnic diversity within a municipal-level electoral district. However, I cannot reject the possibility that ethnic diversity does have an independent effect on party system outcomes. Evidence from the 2004 election suggests both rent opportunities and municipal diversity matter. Hypothesis 4: the effect of ethnic diversity in the absence of rent opportunities The Indonesian case offers a rare opportunity to test the Hypothesis 4. Indonesia’s upper and lower houses are elected using distinct rules which lead to distinct predictions for the competing models. Candidates running for Indonesia’s upper house cannot affiliate with political parties and are not formally tied to the partisan networks that control sub- national rents. As such, we can consider upper house elections to take place in a context with low rent opportunities. A correlation between the upper-house electoral fragmentation and ethnic diversity would strongly corroborate the Communal Voting model. To test Hypothesis 4 I calculate the effective number of electoral candidates (ENEC) for each DPD district in both 2004 and 2009. ENEC ranged from a low of 3.2 to a high of 30.3. The key independent variable is the ethnic fractionalization of the DPD district. Results appear in Table 18. Model 1 presents a stripped down test that includes only the ethnic diversity variable and dummy variable for the 2009 election to control for change across time. The ethnic fractionalization variable does not come close to reaching standard levels of statistical significance. The absence of a clear relationship is 198 demonstrated in Figure 27, which plots the relationship between ethnic fractionalization and the effective number of electoral candidates. The non-relationship is consistent in both 2004 and 2009.225 Table 18 – Determinants of DPD electoral fragmentation Determinants of Party System Size – Upper House (DPD) Election (OLS Regression) Dependent Variable: Effective Number of Electoral Candidates Model 1 – Simple Model Model 1 – Full Model Ethnic Fractionalization 0.30 (2.71) 2.03 (1.86) Number of Candidates (logged) 10.33*** (1.19) Urbanization -7.54** (2.97) Year 2009 1.52 (1.50) -.05 (1.06) Const 15.48*** (2.11) -16.77*** (4.44) Observation 65 65 R2 0.0167 0.5768 Figure 27 - Ethnic diversity and electoral fragmentation in DPD elections 225 See Appendix G, Section 8. 199 In Model 2 I add two additional control variables. First, I include number of candidates. The raw number of candidates ranged from 8 to 69. More competitors should have fragmented the vote. I log the variable as I expect adding one more candidate to a small pool of contenders should have more of an impact than adding an additional candidate to a large pool. Second, I add a variable capturing urbanization. At least in the 2009, urban voters tended to coordinate on fewer partisan options for lower house elections and I anticipated a similar dynamic in the upper house. Even when the controls are added, ethnic fractionalization does not reach statistical significance in any sample. Urbanization is negatively signed and statistically significant. The key independent variable in every model is the logged number of candidates. Increasing logged number of candidates by one standard deviation add 4.5 more effective DPD candidates. The one variable alone explains over 50% of the variance in ENEC. There is no evidence to suggest that ethnic diversity leads to a fragmented electoral vote in DPD elections and thus there is evidence to support Hypothesis 4. The result is surprising in light of the literature on ethnic campaigns. Given that DPD candidates are unable to promote themselves using the party labels, we might expect them to rely instead on appeals recognizable identity cues like ethnicity. First hand observation confirms that candidates do use ethnic cues background to promote their campaigns.226 More finely grained data might reveal links between a candidate’s support base and her ethnic identity. Yet there is no sign that voters coordinate their support with enough discipline to produce ethnically derived patterns of electoral fragmentation. Likewise, the tight correlation between the number of candidates and the fragmentation of the electoral vote indicates that voters do not feel compelled to coordinate around front-runners. As of 2009, elections to the DPD have not been the scene of pitched ethnic competition for supremacy. The findings are more consistent with the rent opportunities model than the communal voting model. In the rent opportunities story, local rewards entice viable elites to enter the competition. Once in the race, elites from minor parties are able to win votes because locally oriented voters do not feel compelled to support major parties. In DPD 226 See Batubara’s skilful use of ethnic cues in Chapter 6 (“Personal Vote”). 200 elections, however, the crucial linkage between local rents and candidate entry is severed. Whereas a strong DPR candidate can provide a locally oriented rent-seeking network with down-ballot coat-tails, these networks do not feel compelled to run friendly candidates in DPD elections. Consequently, high diversity / high rent electoral districts do not witness electoral fragmentation. Conclusion This chapter presents strong evidence to support the rent opportunities model. Within national electoral districts a correlation between rent opportunities proxies and electoral fragmentation does exist. Examining sub-district dynamics reveals that rent opportunities lead to fragmentation in ethnically homogenous areas. When an electoral district is fragmented, it is fragmented in both ethnically homogenous and ethnically diverse municipalities. A study of upper-house results demonstrates that ethnic diversity does not lead to electoral fragmentation in the absence of rents. Empirical evidence, then, supports Hypotheses 1 and 3 but not Hypothesis 4. Despite this evidence, there remain signs of communal voting. Ethnic fractionalization does correlate with party system size in national electoral districts. In fact, the ethnic fractionalization variable explains more variation than the rent opportunities proxy. Diverse municipalities do have high levels of party system fragmentation even when controlling for rents. Consistent with previous research, there is even evidence of an interactive effect between district magnitude and ethnic diversity. Almost all the major observable implications of the rent opportunities theory are present; nonetheless, the ethnic fractionalization measure strongly correlates with district- level party system size even when proxy variables for rent opportunities are added to the statistical model. It is possible that the mixed results are a product of measurement error. The correlation may driven not by the direct effect of ethnic voting but by an indirect association with some other causal process. Given the evidence that lines up with the Rent Opportunities model, the safest conclusion is that both rent opportunities and ethnic diversity are capable of producing an expansion of the party system. 201 Chapter 8 – Party System Change The Indonesian party system has expanded with each successive legislative election. Between 1999 and 2009, the effective number of electoral parties at the national level increased by 4.5, from 5.1 to 9.6. A similar growth took place at the district level, where the average number of parties increased by 4, from 4.5 to 8.5. In effect, the party system almost doubled in size in a ten year time span. The expansion is particularly curious given that it took place in the context of an increasingly restrictive electoral system. What caused the expansion of the party system? This question has recently been taken up by Choi (Choi 2010). Choi makes a two-pronged argument about party system size. First, he argues that neither social cleavages nor changes in the electoral system can account for the change. Second, he suggests the expansion may be caused by the introduction of presidential elections and the rise of ‘new political issues.’ In Choi’s account, the introduction of executive elections prompted aspiring presidential candidates to form new parties. Similarly, increased public concern with corruption facilitated the growth of parties that win votes based on their anti-establishment image. Choi attempts to account for district-level phenomena with national-level changes. However, between 1999 and 2009 the expansion of the party system has been noticeably uneven. In the province of East Nusa Tenggara, the effective number of parties increased from 3.1 to 12.2, while in Jakarta it climbed from 4.5 to 5.6. In Aceh the effective number of parties actually went down (from 6.3 to 5.1). We cannot overlook the fact that the evolution of the party system has taken substantially different paths in different districts. This chapter links the change in party system size to the dynamic effect of rent opportunities on elite and voter behaviour. Since Indonesia’s inaugural election there have been wide-spread changes in public expectations about the control of sub-national rents. Right up until the 1999 election, many elites and voters expected sub-national state resources to be controlled by one party, much as they were throughout Suharto’s New Order. In areas with plenty of rent opportunities, the traditional party of authoritarian control – Golkar – attracted significant levels of support. Rent opportunities acted as a force of consolidation rather than fragmentation. After years of living with near-universal 202 coalitions, public beliefs about control of rents evolved. In 2004 and 2009, candidates and voters expected elected legislators to form over-sized coalitions that would distribute rents widely. There was little incentive to line up behind a major party. As a result, sub- national party systems that had previously been held together by the promise of state rewards fractured. The argument provides an answer to one the unresolved questions of the dissertation: why did ethnicity and rent opportunities not correlate with party system size in 1999?227 What made the 1999 election different? My argument proposes that the effect of rents is conditioned by patterns of rent sharing. The correlations between party system fragmentation and rents/ethnic diversity that we observe in 2004 and 2009 occurred only after the expectation of rent sharing solidified. Before that point, individual parties were able to credibly claim an ability to control sub-national rents. The chapter also addresses the key oversights in Choi’s argument. Whereas Choi proposes that the evolving politics of corruption have played an important role in party system change, he does not identify a specific mechanism connecting the issue to party system fragmentation. Consequently, he is unable to account for regional variation in party system expansion. This chapter proposes a fleshed out mechanism connecting local rent opportunities to district-level party system size. The chapter proceeds as follows. First, I review the literature on party system change in Indonesia, focusing closely on Choi’s recent piece. Second, I offer a theory connecting evolving beliefs of rent sharing to changes in elite and voter behaviour. Third, I demonstrate the consolidating impact of rents in 1999 through an examination of municipal-level party systems in that year. As well, I connect party system expansion to rent opportunities. Fourth, I place party system change in the context of Golkar’s decline. Indonesian party system change: existing explanations Electoral institutions and ethnic diversity: the dogs that did not bark Choi’s analysis of party system change in Indonesia focuses closely on the factors that he believes are not causing fragmentation. First, he asserts that electoral system change is 227 For evidence of the non-correlation between ethnic diversity and party system fragmentation in 1999, see Appendix H, Section 1. 203 not the culprit. It is often held that district magnitude is the decisive factor in determining party system size (Cox 1997; Reed 1990; Taagepera and Shugart 1989). The number of seats sets an ‘upper-bounds’ on the number of parties that can exist in equilibrium. This ‘M+1’ rule (where ‘M’ is district magnitude) does not predict how many parties will necessarily exist, but it does suggest low magnitude districts should have fewer parties. Yet, as Choi notes, district magnitude has decreased each election in Indonesia yet party system size has increased. In fact, party system size is commonly higher than the ‘M+1’ upper bounds. Factors other than district magnitude are shaping the party system. Choi also challenges the idea that social diversity is driving party system change. Shrinking district size has meant the homogenization of electoral districts.228 Having noted shrinking district size, Choi draws the conclusion that social diversity could not possibly be responsible for increasing party system size, a statement he declares true “no matter how we measure or define social diversity here” (Choi 2010, 677). Relying on the same logic, he discounts the idea that an interactive relationship between social diversity and district magnitude exists. Choi’s argument relies on two implicit assumptions: 1) the effect of ‘social diversity’ works through the mechanism of communal voting; 2) the effect of social diversity is uniform across elections. On the first, Choi simply follows the existing literature on ethnic diversity and voting. He assumes identities (ethnic or otherwise) influence vote choice and communities vote as an ethnic block for a distinct party or parties. This orthodox account overlooks the potential for indirect diversity effects. The second assumption also conforms to the prevailing literature but is noteworthy for two reasons. First, Choi himself acknowledges that politics during transition were indicative of post-transition patterns. His account of party system change relies on the rise of ‘new’ issues in the post-transition period. The same logic can be applied to the potential effects of social diversity. Second, Choi does not investigate the effect of social diversity empirically.229 If he had, he would find that the most prominent 228 Though Choi does not measure this, his intuition is correct. The average largest ethnic group within a district grew from 45% in 1999 to 63% in 2004. Interestingly, largest average ethnic group size held steady at 63% between 2004 and 2009 despite the addition of 8 new electoral districts. 229 To explain why he did not investigate any diversity measures, Choi cites lack of data. He notes, “The Indonesian government publishes relevant population data on parameters such as religion and ethnicity 204 measure of diversity – ethnic diversity – correlates with party system size in 2004 and 2009, but not 1999.230 This confirms the intuition that there is something different about electoral politics in 1999. It also suggests the effect of diversity evolved over time. Presidentialism and new issues: explaining change Choi’s own explanation emphasizes the role of executive elections and the rise of new issues, namely corruption. First, the introduction of direct presidential elections in 2004 changed the way presidential candidates interact with the party system. The presidential contest contributed to the personalization of Indonesian campaigns. Aspiring presidential candidates believe they can win without the support of an established party.231 However, given electoral laws that require presidential candidates to reach a minimum benchmark of electoral or legislative support, aspiring executive candidates have a strong incentive to launch their own parties to support their bid. This process has undoubtedly contributed the process of fragmentation. In addition to the introduction of new presidential parties, Choi also asserts party system change is driven by “The rise of new pressing issues (including corruption), which are not the same as the established social cleavages but cut across them” (Choi 2010, 681). In this account, a growing proportion of voters now strongly oppose corruption and rally to parties that can credibly claim to fight the system. Choi’s causal story relies on the assumption that all parties are perceived as corrupt and thus voters must look beyond the established system to find an ‘anti-corruption’ option.232 Beyond the unstated assumption - which is consistent with Slater’s popular account of the Indonesian party system - problems exist with the timing of Choi’s story (Slater 2004). only at the provincial level” (Choi 2010, 675). It is true that we cannot track changes over time using only the 2000 census data. Yet municipal data from 2000 is contained within the census publications. Furthermore, the ethnic fractionalization scores derived from the more readily available provincial-level data demonstrate a similar statistically significant correlation only after 1999. 230 For 1999 results, see Appendix H, Section 1. For 2004 and 2009 results, see Chapter 7. 231 This was arguably true in 1999 as well. The formation of both PAN and PKB were closely tied to the presidential ambitions of their respective leaders (Amien Rais and Abdurrahman Wahid). These two parties tend not to be treated as mere presidential vehicles as their support was drawn from pre-existing religious organizations. This strong civil society connection gives the parties a ‘rootedness’ not found in later presidential parties. 232 Choi comes close to recognizing this assumption. Elaborating on the potential mechanism, Choi states “we may say that widespread popular discontent with the democratic government’s unsatisfactory performance is one of the key factors in the transformation” (Choi 2010, 680). 205 An anti-party/anti-incumbent reaction is a compelling explanation in 2004, but hardly in 2009.233 Between 2004 and 2009 the President’s party increased its national vote share almost 300%. The new modestly sized parties (Hanura and Gerindra) were both led by highly compromised figures (Wiranto and Prabowo) that had no standing as anti- corruption activists. If anything, the growth of ‘anti-corruption’ parties (PD and PKS) had a consolidating impact on the party system between 2004 and 2009. The source Choi relies on to establish the increasing importance of corruption actually forwards a distinct mechanism more in keeping with this dissertation. According to Peter King (2008), during the New Order “[corruption] was closely controlled by one authoritative figure at the Centre and therefore less messy and pervasive than the looser, more decentralized and competitive, corruption we have come to know since reformasi [reform]” (P. King 2008). The break-up of old networks and the rise of new actors was facilitated by the transfer of authority to sub-national units. As King notes, “decentralization of government powers and budgeting to provinces and regencies has seen a new class of prosperous public and private sector corruptors appear at the local level” (P. King 2008). King’s account of shifting networks of corruption suggests an entirely different mechanism connecting corruption and party system fragmentation than that forwarded by Choi. Rather than a struggle against corruption fragmenting politics, competition tends to involve a multitude of loosely connected actors seeking control of sub-national resources. Rent opportunities, rent sharing, and party system change To understand variance in district level party system expansion we need to look at the factors that motivate political competition. It is my contention that sub-national rent opportunities shift the locus of political activity from the national to the local level. Where the state plays a significant in the local economy, a persons livelihood is directly impacted by who controls public office. For elites, there is a strong incentive to enter politics so as to control state resources, either for personal enrichment or simply to enhance community standing. For voters, there is a strong incentive to connect oneself to a powerful benefactor who can provide access to state resources. Whereas voters and 233 On the anti-party reaction, see Johnson Tan (Tan 2002). 206 elites in low rent areas may be motivated by national leadership concerns and/or broad policy issues, political actors in high rent areas focus more on the control and distribution of sub-national state resources. Machine politics or partisan melee? The effect of rent sharing expectations The impact of rent opportunities on the party system depend upon prevailing expectations of rent sharing. Rent sharing refers to the existence of universal or near universal coalitions in which legislators can access rents. This does not mean there is complete equality in rents accessed; some legislators in the coalition may enjoy access to more rents than others. Rent sharing does imply the existence of a minimal threshold of rent access that separates ‘ins’ from ‘outs.’ Rent sharing expectations refer to ex ante beliefs held by elites and voters regarding the distribution of state rents. There is an expectation of rent sharing if voters and elites expect universal or near universal coalitions to form after an election. On other hand, there are no expectations of rent sharing if elites and voters anticipate a single-party or a minimal winning coalition will monopolize access to rents after an election. The causal pathway leading from high rent opportunities to a high number of parties depend upon rent sharing expectations. Where rent opportunities are high and expectations of rent sharing exist, aspiring politicians join minor parties in large numbers because they expect even minor parties can access a healthy portion of local rents. Voters, for their part, find the promises of favours from minor party candidates credible because they expect minor parties to participate in large coalitions. The theorized effects that rent opportunities have on candidate and voter decision-making hinge on the underlying beliefs that actors hold about minor party power and post-election coalition politics. Minor parties are reliable vehicles for attaining power when universal coalitions are the norm, but they have a harder time attracting support when political actors expect the major parties to monopolize state resources. Where there are expectations that rents will be controlled by a major party or a minimal winning coalition, the relationship between rents and party system fragmentation is transformed. Accessing rents remains a top priority in high rent areas, but the methods of accessing rents necessarily shift. In this context, few elite sign up for minor parties and few voters lend them their support. 207 Instead, political attention gravitates toward the large parties that are most likely to dominate government formation. A large party that can commit to distributing rents will benefit from low rent sharing expectations. There are different ways a party can establish its credibility as a party machine.234 It can recruit candidates known to distribute rents. It can signal its attention to distribute rents through its platform. Or it can rely on a reputation built through a periods of government control. The party that can most credibly commit to distributing state resources to supporters will benefit in areas where rent opportunities are high. In sum: the effect of local rents depends on expectations of rent sharing. When elites and voters expect rent sharing, we see processes that fragment the political system in high rent areas. But when there is an expectation that rents will be controlled by one party, there are strong incentives to coordinate activity within a political machine. As such, rents consolidate the party system in those areas. Rent sharing in post-Suharto Indonesia Since the fall of Suharto, expectations of rent sharing have evolved. In the lead-up to Indonesia’s first post-Suharto election, expectations of rent-sharing were low. President Suharto had long dominated Indonesian politics and sat at the top of a patronage pyramid that extended down into the sub-provincial units. The largest party – Golkar – enjoyed a long history of privileged access to state resources. From the point of view of the voter, the structure of authority looked much like it used to, though elections were now freely contested. For those in high rent locales, the most practical strategy appeared exactly as it always had: line-up behind the hegemonic party. Compared with low-rent areas, party systems were relatively concentrated. Reform of the formal institutions, combined with changes in informal rules guiding party interaction, led to expectations of local rent sharing in 2004 and 2009. Rent opportunities impacted party system size in distinct ways in the two different time periods. Low expectations of rent sharing in the lead up to the 1999 election had a consolidating effect on party system size. Voters in high rent sub-national systems 234 By ‘machine’ I refer to an impersonal organization that exchanges state favours for votes. 208 expected the previously dominant party machine – Golkar – would continue to control state resources. Accordingly, the number of parties in high rent areas was low. This narrative points toward three empirical hypotheses that can be tested statistically: H1: The number of parties is negatively associated with rent opportunities in 1999. H2: The expansion of the party system between 1999 and 2009 is positively associated with rent opportunities H3: The percentage of Golkar voters is positively associated with rent opportunities in 1999 These three hypotheses test distinct aspects of the causal story. H1 proposes that rents had a consolidating effect on party systems in 1999. Following on this logic, H2 holds that the expansion of the party system was directly tied to shifts in the way voters and elites perceived rent distribution. H3 fleshes out the mechanism by connecting the strength – and subsequent decline – of the Golkar machine to its ability to control sub- national rents. Theory testing: change across time Hypothesis 1: municipal party system size in 1999 To test Hypothesis 1, I used results aggregated by municipality. Exploration of electoral dynamics across national districts was limited in 1999 due to the small number of observations (26) but there was plenty of variation across the country’s 312 municipal units. The primary dependent variable is the effective number of parties within a municipality. Votes from the national-level election contest were used to calculate the municipal ENEP scores. I relied on results collected by the International Foundations for Electoral Systems. The mean municipal ENEP was 4.05; the standard deviation was 1.29. The key independent variable is civil service size, a proxy for rent opportunities. Here I use the proportion modern sector workers employed in the civil service. Data was drawn from the 1999 Indonesia in Statistics Yearbook. Given data restrictions, all municipalities within a province are assigned the same value. Also, I included data on the district’s ethnic structure. These were constructed using data from the 2000 census. Reconstruction of municipalities was necessitated due 209 to the splitting of districts. I was able to generate scores for 311 of 312 districts. The mean fractionalization was 0.42, with a standard deviation of 0.33. I added two key control variables. The first was a simple dummy indicating whether or not the municipal unit is a ‘city’ (kota). This serves as a rough measure of urbanization. The previous chapter found a positive relationship between urbanization and party system size in 2004 and a negative relationship in 2009. I make no predictions about the direct of the relationship in 1999. Second, I added the logged district magnitude of the national electoral district. Results from a three variable model appear in Table 19. In Models 1 and 2 I test for independent effects of ethnic diversity and civil service size. Consistent with Hypothesis 1, there is a negative correlation between civil service size and the number of parties in Model 1. The correlation is significant at the P<.01 level. As shown in Model 2, there is no statistically significant relationship between ethnic fractionalization and municipal party system size. It is, however, positively signed.235 Table 19 – Party system size [1999] Determinants of Party System Size – National Legislature (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 Model 2 Model 3 Model 4 Civil Service Size -0.05*** (0.01) -0.09*** (0.02) -0.05 (0.04) Ethnic Fractionalization 0.44 (0.30) 0.71** (0.30) 1.42** (0.67) District Magnitude (logged) -0.29*** (0.10) -0.02 (0.10) -0.29** (0.11) -0.24** (0.12) City 0.28 (0.17) 0.31 (0.19) 0.18 (0.19) 0.15 (0.19) Civil Service X Ethnic Fractionalization -0.07 (0.06) Const 5.37*** (0.42) 3.93*** (0.38) 5.56*** (0.53) 5.05*** (0.68) Observations 312 285 285 285 R2 0.0490 0.0269 0.0882 0.0928 In Model 3 I add both key independent variables. Both variables are strongly significant. Civil service size remains negatively signed, while ethnic fractionalization 235 Note: It is difficult to determine whether municipalities where the largest group is ‘other’ are dominated by one large group or a collection of small groups. Consequently, I exclude all municipalities where the largest group is ‘other.’ 210 remains positively signed. All else being equal, moving from the minimum to the maximum civil service size results in a reduction of 2.55 electoral parties, while a similar change in ethnic fractionalization produces an increase of 0.68 more electoral parties. These are substantively large effects, especially the reductive impact of civil service size. To illustrate the magnitude of the effects, I used Clarify to run generate party system predictions for municipalities from Lampung and South Sulawesi. In 1999, approximately 8.48% of non-agricultural workers in Lampung were employed in the civil service, which put municipalities in the province close to the mean civil service size of 10.9%. In contrast, 14.09% of non-agricultural workers in South Sulawesi were employed as civil servants. The difference in civil service size between the two provinces (5.61%) was close to one standard deviation in civil service size (5.98). Comparing predicted values across the two provinces allows a rough comparison in the effect of the key rent opportunities proxy variable. For each province I generated predictions in two municipalities, one with an ethnic fractionalization score close to the mean (0.42) and another about one standard deviation above the mean (0.75). Results appear in Table 20. In each province the model predicts a modest effect of ethnic fractionalization. Moving from relatively homogenous East Lampung to relatively diverse North Lampung yields an increase of 0.25 electoral parties, while moving from relatively homogenous Pinrang to relatively diverse Luwu produces a predicted increase of 0.35 parties. A more substantive effect is observed in the cross-province comparisons. The model predicted that the relatively diverse municipality of Luwu would have 0.61 fewer electoral parties than the diverse district of North Lampung. Likewise, homogenous Pinrang was predicted to have 0.71 fewer parties than homogenous East Lampung. This finding indicates that rent opportunities likely had a strong consolidating effect on the party system in 1999.236 236 There is weak evidence for a conditional relationship. Rent opportunities acted as a ‘brake’ on the centrifugal effect of ethnic diversity. For further discussion see Appendix H, Section 2. 211 Table 20 – Predicted municipal party system size [1999] Province Municipality Ethnic Fractionalization Civil Service Size (% non- agricultural) Predicated ENEP Lampung Lampung Timur 0.38 8.48 4.26Lampung Utara 0.73 8.48 4.51 Sulawesi Pinrang 0.28 14.09 3.55Luwu 0.77 14.09 3.91 In sum: Evidence from the section is consistent with Hypothesis 1. The collected findings both support and challenge the argument of the dissertation. At the level of the municipality, ethnic diversity correlates with the fragmentation of the electoral vote. This strongly suggests that ethnic diversity has an independent effect on party system size. Yet there are clear signs that rent opportunity also consolidated the party system in that year. Indeed, the magnitude of the rent opportunities variable was substantially larger than the effect of the ethnic diversity variable. The finding helps explain why we do not see the relationship between party system size and rents/ethnic diversity in 1999. Hypothesis 2: district level party system change over time To understand the 2009 district-level results it is essential to consider patterns of regional change over the ten-year-time span. To track party system change through time I reconstructed the 2009 electoral districts using results from the 1999 national election. For each district I then calculated the reconstructed ENEP. The mean reconstructed district had 4.2 parties, with a standard deviation of 1.6. In 2009 the mean district had 8.5 parties, just over twice the number of effective parties from 1999. Comparing district-level party fragmentation scores in 1999 and 2009, we find the two variables have a correlation coefficient of 0.006 with a standard error of 0.20.237 This does not even approach standard levels of statistical significance. The non-correlation makes sense in light of the municipal results. In 1999 diverse districts contained counteracting forces, one centrifugal (many ethnic groups) and one centripetal (rents). By 2009, both of these forces were working in the same direction causing severe party system fragmentation in high diversity/high rent districts. The non-significant relationship between party system size in 1999 and 2009 clearly indicates that the causal 237 For Figures, see Appendix H, Section 3. 212 processes producing party system fragmentation in each of these two elections was distinct. The most fragmented districts in 2009 were the ones that experienced the most expansion across the decade.238 This is intuitive, but other potential evolutionary paths existed. For example, expansion could have occurred at an even rate, or the most fragmented districts with the most parties could have become even more fragmented than the rest. Instead, cross-district variation in the number of parties in 2009 reflected an uneven process of party system expansion. The roots of contemporary party system size lay not in the patterns of 1999 but rather the changes that occurred between the elections. To uncover the correlates of party system change I created a party system expansion variable. This was done simply by subtracting the effective number of parties in 2009 by the corresponding effective number of parties in 1999. The mean party system expansion was 4.3, with a standard deviation of 2.2. There was a significant range of scores, from a high of 10.3 to a low of -1.7. I used party system expansion as a dependent variable in a multivariate regression. To construct the model I added many of the same variables used in the party system size model from Chapter 7. These included urbanization, an Aceh dummy, ethnic fractionalization, and civil service size. Table 21 presents results. The two control variables – urbanization and Aceh – are both negatively signed and statistically significant. Unsurprisingly, the special laws in Aceh that allowed for the formation of regional parties helped consolidate the party system in that province. The negative relationship between expansion and urbanization indicates the development of a distinctive urban political dynamic led by President Yudhoyono’s PD and the Islamist leaning PKS. The anti-corruption, anti-establishment messages of both parties tended to resonate with educated voters of the urban middle- class. This particular upwardly mobile constituency has greater private sector opportunities and their fortunes are less dependent on their connections with state officials. In 2004, both parties began making their breakthrough in urban areas. In 2009, 238 See Appendix H, Section 3. 213 the two parties consolidated their urban gains in 2009. 239 In the process, the rise of the urban parties consolidated local party systems. Table 21 –Determinants of party system expansion Determinants of Party System Expansion, 1999-2009 – National Legislature (OLS Regression) Dependent Variable: Change in Effective Number of Electoral Parties Model 1 Model 2 Civil Service Size 0.06 (0.05) -2.75*** (0.60) Ethnic Fractionalization 1.13 (0.74) 1.89** (0.77) Urbanization -4.40*** (0.87) -4.05*** (0.78) Aceh -6.77*** (1.15) -5.24*** (1.09) Corruption Index [TI] -5.57*** (1.23) Corruption X Civil Service 0.52*** (0.11) Const 5.15*** (0.54) 34.63*** (6.53) Observations 77 77 R2 0.5144 0.6314 Ethnic fractionalization and civil service size are both positively signed, but neither variable is statistically significant. This is likely a problem of multicollinearity; when the model is run excluding one of the variables, the remaining variable is positively signed and statistically significant.240 To parse out the independent effects of rents and ethnic diversity I refined the rent opportunities measure. Civil service size is simply one proxy of rent opportunities that captures the resources available to politicians. I added to the model a measure of ‘constraints’ on elite behaviour. Here I relied on Transparency International Corruption Perception data. Unfortunately we only have data in two years (2008 and 2010). I use this limited data to construct a provincial-level corruption averages. I subtracted the average 239 For instance, the combined vote total of the two parties was a majority in five electoral districts, all major urban centres: 1) West Java 1 [Bandung]: 55.8%; 2) Jakarta 1: 55.3%; 3) Jakarta 3: 52.2%; 4) West Java 6 [Bekasi & Depok]: 50.7%; 5) Jakarta 2: 50.6%. 240 See Appendix H, Section 4. 214 from 10, so that the corruption score increases as perceptions of corruption increase. This ‘constraints’ proxy was interacted with the existing ‘resources’ proxy to generate an interactive term that more accurately captures the rent opportunities concept. Table 21, Model 2 contains results. In order to interpret results, Figure 28 replicates Brambor et al.’s procedure of measuring conditional effects (Brambor, W. R. Clark, and Golder 2005). We see a clear relationship: a one unit change in corruption leads to a high number of parties only in districts with large civil service. In a province with a moderately sized civil service (10% of the modern sector), an increase in corruption does not correlate with party system change; however, in a province with a relatively large civil service (20%), an increase in one unit of corruption correlated with an increase of 5 electoral parties.241 Figure 28 – Rent opportunities and party system expansion To isolate the potential effect of rent opportunities it is useful to compare predicted values of electoral districts with similar ethnic structures. Returning to the province of South Sulawesi, the electoral district of South Sulawesi 3 has a moderate level of corruption (5.7) but a large state sector (13.7). The electoral district of Lampung 1 has a similar corruption score (5.2) but a modest civil service size (7.99). These districts 241 The change appears abnormally large; however, the variation in the corruption index score is modest. The minimum value is 3.8, the maximum 6.4, and the standard deviation 0.4. 215 contrast with West Java 8, which shares similar corruption scores (5.5) but resides in a province with relatively small civil service (3.8). All three districts, however, have moderately fragmented ethnic structures.242 To generate predicted values I used Clarify. For the purposes of the simulation I set Aceh to 0, the urbanization variable to the mean, and the ethnic fractionalization value to 0.7, the average fractionalization of the three districts. Predicted values appear in Table 22. The model predicts the highest change in party system fragmentation (5.5) in South Sulawesi 3, the electoral district with the most resources state available for manipulation by political elites. Lampung 1, with a moderate civil service, produces a moderate change in party system fragmentation (4.9). West Java 8, the electoral district in the province with the smallest civil service, produces the smallest change in party system size (4.1). Thus controlling for ethnic fractionalization across electoral districts with similar level of constraints on elite behaviour reveals the independent effect of rent opportunities. Table 22 – Predicted party system expansion District Name Corruption Score Civil Service Size (% non-agricultural) Party System Expansion West Java 8 5.5 3.8 4.1 Lampung 1 5.2 8.0 4.9 South Sulawesi 3 5.7 13.7 5.5 The findings from this section indicate that the electoral districts that experienced high levels of party system expansion tended to have both a large civil service and a high level of perceived corruption. This closely conforms to the rent opportunities story. In short, the evidence from the section supports Hypothesis 2. Hypothesis 3: The decline of Golkar and party system change The story of Indonesia’s expanding party system is closely connected to the story of Golkar’s decline. Golkar is the only major party from 1999 where the percentage of votes lost in the last decade positively correlates with the expansion of the party system.243 This 242 Ethnic fractionalization scores: South Sulawesi: 0.77; Lampung 1; 0.70; West Java 3; 0.63. 243 See Appendix H, Section 4. 216 is not to say that the changes in other parties did not have an important impact.244 The decline of the other ‘Big-5’ parties, however, did not result in the same magnitude of party system change. Looking closely at the decline of Golkar allows us to track the structural causes of party system expansion. The End of the ‘Monoloyalitas’ Era Golkar was Suharto’s party of authoritarian control. Suharto used Golkar to organize massive electoral victories in a series of controlled elections that took place between 1971 and 1997. The party was effective for several reasons, but one important factor was its ability to control access to state resources. Career prospects of many educated Indonesians were tied directly to Golkar fortunes through the ‘mono-loyalty’ (monoloyalitas) regulations which required all civil servants to be part of Golkar. Golkar membership was also a prudent policy for businessmen interested in securing state contracts. For a large segment of the Indonesian population interested in upward mobility, careers were advanced through Golkar participation. The government repealed the mono-loyalty laws prior to the 1999 elections. While the repeal of the law officially detached Golkar from the state bureaucracy, many bureaucrats and would-be politicians looking to maintain control over state resources stayed in the party. This was particularly true in areas where the state played a substantial role in the local economy. As noted by van Klinken (2007), Golkar’s provincial electoral vote in 1999 tended to closely follow the province’s civil service size.245 This strongly suggests that Golkar was able to use its traditional dominance of state patronage to secure electoral support in 1999. Following the 1999 elections, Golkar’s grip on the bureaucracy weakened. The most obvious sign of Golkar’s official decline could be seen in the ascension of two ‘reformasi’ parties to the Presidency (Wahid/PKB) and Vice-Presidency (Megawati/PDI- 244 Perhaps the most powerful fragmenting factor not explicitly dealt with in the rent opportunities model was the break-up of PKB. While the party won only 12% of the national vote in 1999, it managed to dominate many municipalities in East Java where it relied on the religious organization NU to mobilize supporters. Intra-party conflict contributed to a precipitous electoral decline in the 2009 legislative elections. The areas where it experienced its sharpest declines were also those that experienced considerable increases in party system size. For instance, in three districts – East Java 2, 3, and 11 – PKB lost over 30% of the electoral vote. These three districts also experienced the most significant party system expansions in all of Java (6.16, 6.49, and 6.71). 245 My own calculations confirm van Klinken’s (van Klinken 2007). 217 P). Additionally, coalition politics made it increasingly difficult to separate winners from losers. At the national-level, Wahid’s first post-election cabinet enveloped all major parties. Following Wahid’s impeachment, Megawati revived the grand coalition. Even at the local level, over-sized coalitions and power-sharing became the norm. These power shifts signalled that ambitious politicians and bureaucrats no longer required Golkar service to get ahead in their careers. While this reduced incentives to stick it out within the party, there was no obvious successor party machine to take its place. Factionalism and Inflexibility in the Reformasi Era In addition to the party’s loosening grip on power, a modernizing faction within Golkar attempted to change the fundamental nature of the party. The internal tension revealed itself in the battles between the ‘Iramasuka’ faction and its enemies, particularly those grouped around Golkar Chariman Akbar Tandjung.246 Iramasuka was a loose collection of Golkar functionaries from Eastern Indonesia. In 1999 they supported Jusuf Habibie and, in the lead up to the 2004 presidential race, backed Wiranto over Tandjung. Their power rested on their ability to control government positions in the patronage rich regions of the East. They were largely successful in this goal; in fact, in 1999 Golkar gained slightly over 40% of the vote in Eastern provinces. Figure 29 plots the Golkar vote by region. In 1999, the party’s percentage of votes in the East was over double its vote percentage in Java and considerably higher than the total in Sumatra. Iramasuka could reasonably claim to represent what was left of Golkar’s voter base. 246 ‘Iramasuka’ stands for IRian,MAluku, SUlawesi, and KAlimantan. In Indonesian, the acronym also means ‘happy melody.’ For a compherensive look at the faction, see Tomsa (Tomsa 2006). 218 Figure 29 – Golkar support by region In Jakarta, Iramasuka fought what was often a rear-guard defence against both internal and external enemies. Their championing of Habibie, the Sulawesi-born interim President, demonstrated not just a simple regionalism but a resistance to move beyond New Order figures. In addition, Iramasuka staunchly backed cohorts like Arnold Baramuli and Timmy Habibie, who were both implicated in the embarrassing Bank Bali scandal. The faction was, in many respects, a product and defender of the Golkar’s bureaucratic machine. 247 Tandjung’s power base lay among the Javanese functionaries.248 Despite gaining nowhere near as high a percentage of voters on Java as compared to Eastern Indonesia, the largest bloc of votes for the party still came from the densely populated island. Tandjung was hardly a hardly a clean figure and his zeal for reform was greatest when it happened to align with his own prospect of career advancement.249 But he and his allies were mindful that Golkar would need to modernize its image if it wanted to compete in the new, open electoral environment (Ziegenhain 2008, 86). Among other things, Tandjung orchestrated the ouster of Habibie and provided political cover for Marzuki 247 O’Rourke referred to Baramuli, a founder of Iramasuka and former governor of South Sulawesi, as “grossly at odds with the reformasi era” (O’Rourke 2002, 222) 248 Tandjung himself was born in North Sumatra. 249 Habibie, for instance, was more willing than Tandjung to act as reformer when it came to crafting the civil service law (D. Y. King 2003, 63-6). Golkar Support by Region 0 10 20 30 40 50 1999 2004 2009 Election Year Pe rce nta ge G olk ar Vo te Sumatra Java East 219 Darusman’s anti-corruption campaign, moves which provoked opposition within the party. These reform measures allowed Golkar to place some distance between itself and its New Order image. Though Tandjung’s sack of Habibie may have provided Golkar greater opportunity to reinvent itself, it also sapped the internal strength of the party. The efforts of the ‘modernizing’ faction, particularly the targeting of prominent politicians from the East, went some ways toward alienating the party from its core supporters. The decline of Golkar in the East was a likely outcome in any case, but it was hastened by factional politics. Beyond the factional tensions, the internal rules and structures hindered Golkar’s shift from an authoritarian to democratic political party. During the Suharto period, Golkar placed a premium on loyalty to the organization. Candidates for office were expected to have a history of party service. By forcing members to invest considerable time and energy before becoming a trusted Golkar cadre, the party was able to socialize recruits into party rules and norms. Aspects of the rigid structure were partially carried into the reformasi period. Even in 2009, candidates for national legislative office were still expected to have 5 years service as a Golkar member, a formal requirement that went considerably beyond any of its competitors. During the New Order Golkar could depend on its privileged access to patronage to attract new talent. After the transition, Golkar had to compete on a more level playing field. While all rules in Indonesia can be bent for the right price, Golkar’s demanding structure deterred aspiring politicians from joining the party. As the intake slowed, the outflow quickened. Sunk cost investments and years of socialization kept many existing Golkar members within the party structure, yet a sizable number of Golkar stalwarts were either driven out due to factional struggles or were tempted by alternative party labels. Added to this trend was the natural attrition brought on by retirement, death, and corruption convictions. In short, the Golkar machine was growing old and falling apart. Golkar’s ‘Dead Cat Bounce’ Golkar’s slow demise was masked for a time by the modest success of Tandjung’s modernization project. Golkar won the most votes in the 2004 legislative election. This victory was hardly a case of surging nation-wide support for the party. The percentage of 220 votes between the first and second election actually fell slightly, from 22% to 21%. Nonetheless, Golkar’s decline was much gentler than PDI-P’s, whose electoral vote fell from 33% to 18%. Thus only 5 year after its first defeat in a free election, Golkar again found itself with the largest caucus in the national legislature. Golkar’s relative success however, masked deep internal changes in its voting base. Over 1.9 million new voters in Java were offset by 1.6 million lost voters in the Eastern provinces. This translated into large percentage changes off of Java: whereas their Java vote total crept up from 16% to 18%, Golkar’s Eastern Indonesia vote total dropped from 41% to 30% (see Figure 29). The new, modern Golkar proved an ability to compete in vote rich Java but not without losing its grip on the patronage rich Eastern provinces. Politics in the capital continued on largely independent of the electoral shifts. Only months after the ‘victory’ in the legislative election, the party’s official presidential candidate (Wiranto) lost in the first round of the presidential contest. Always a party to hedge its bets, many Golkar supporters soon found hope in Jusuf Kalla, a Golkar member from South Sulawesi running as Yudhoyono’s Vice-President. Much to the chagrin of Tandjung, who gave the party’s official endorsement to the Megawati-Hasyim ticket, SBY-Kalla proved triumphant. After a quick internal struggle, Kalla was able to use his executive perch to wrest control of Golkar. With Kalla’s take-over, Golkar became the major coalition partner in the SBY government. Golkar’s ascension in Jakarta did not reverse the electoral tide. In the 2009 legislative elections, Golkar lost a further 11% of the vote in Eastern Indonesia, bringing their Eastern total down to 19%. Their Java based bump disappeared as well; between 2004 and 2009, Golkar lost over 5.5 million Java-based voters. Golkar maintained some strength in a few of its traditional strongholds, but it was a fraction of what it used to be. For example, Golkar was only able to gather 25% of the South Sulawesi vote in 2009, down from 66% a decade prior. The story was not uniform decline though. Golkar picked up votes in unlikely places. For instance, in 1999 Golkar’s weakest performance was in the province of Bali, where it scrapped together a mere 10% of the electoral vote. By contrast, the party received 19% of the Balinese vote in 2009, placing the province among the top-5 Golkar 221 boosters. These trends represent the ‘nationalization’ of Golkar; whereas the party’s support used to be heavily concentrated in certain geographic areas, it is now dispersed relatively evenly across the country. Key to the story of the dissertation, however, is the fact that Golkar has not been replaced by a new machine. More often than not, the electoral collapse of Golkar benefitted minor parties. Take Southeast Sulawesi for example. Between 1999 and 2009, Golkar lost 48% of the electoral vote in this district. PD picked up some of these voters, reaching 21% of the total district vote in 2009. However, a majority of the former Golkar voters appear to have spread the support across a large number of minor parties. In 1999, Southeast Sulawesi contained 3 parties that received between 1% and 5% of the vote; in 2009, it contained 19. Golkar’s collapse fractured the system. Measuring the Decline The above story suggests that between 1999 and 2009 Golkar gradually lost the ability to gather votes using state resources. We can see further evidence of this through simple bivariate analyses of Golkar’s vote with the percentage using only the civil servants in the province as an independent variable. Results for all three post-Suharto elections appear in Table 23, Models 1 to 3. In 1999, the relationship between civil service size and Golkar support was strong. For every 1% increase of civil servants in the modern economy, Golkar’s vote increased 1.3%.250 Approximately 23% of the variation in Golkar vote can be explained with this one variable alone. By 2004 we see signs of weakness. For every 1% of civil servants in the modern economy, Golkar’s vote increased 0.5%. This relationship is still significant at the .05 level; however, only 15% of the 2004 variation in the Golkar vote can be explained. The trend is repeated in 2009. For every 1% of civil servants in the modern economy in 2009, Golkar’s vote increased 0.24%. The 250 This result is consistent with van Klinken but at odds with King (D. Y. King 2003, 151-3). The discrepancy occurs due to differences in measurement and modelling strategies. I follow van Klinken in measuring the weight of the civil service in the modern (i.e. non-agricultural) sector whereas King measures as simple percentage of total workers. van Klinken’s strategy is preferable as it more accurately captures the importance of civil service positions to the politically active class. Additionally, King’s models include an odd array of variables that are not theoretically justified. For instance, he discovers that Golkar support in the authoritarian era correlates with Golkar support in 1999, a finding that simply pushes backwards in time the question of why Golkar support was high in some regions and low in others. 222 relationship no longer reaches standard levels of statistical significance and only 9% of the variation can be explained. Table 23 – Golkar support by election Determinants of Golkar Support – National Legislature (OLS Regression) Dependent Variable: % Golkar Votes Model 1 - 1999 Model 2 - 2004 Model 3 - 2009 Model 4 - Pooled Model 5 - Pooled Civil Service Size 1.30** (0.49) 0.52** (0.25) 0.24 (0.15) 0.66*** (0.23) 1.73** (0.83) Election Number -6.82*** (1.35) -0.36 (3.40) Civil Service X Election -0.53* (0.30) Const 13.77** (6.65) 18.06*** (3.54) 13.20*** (2.06) 28.96*** (4.42) 15.83 (9.56) Observations 26 26 26 78 78 R2 0.2261 0.1568 0.0912 0.3387 0.3809 To demonstrate the magnitude of the effect I developed an interactive model of Golkar decline. Again, the dependent variable is provincial Golkar vote. The first independent variable is civil service percentage. The second captures ‘election number,’ which is set to ‘1’ for 1999, ‘2’ for 2004, and ‘3’ for 2009. I expect election number to negatively correlate with Golkar Vote. Third, I add an interactive term. This variable captured the intuition that Golkar’s vote tended to drop the most in those areas with high rent opportunities. The model was run using robust standard errors clustered by province. Table 23, Model 5 contains results. The interactive term is negative and strongly significant. I use Brambor et al.’s method to demonstrate the magnitude of the interactive effect (Brambor, W. R. Clark, and Golder 2005). Results appear in Figure 30. A clear, negative relationship between Golkar vote and civil service size presents itself. Moving from the first to second election in a relatively low rent district (7.5) correlates with a loss of approximately 5% of the Golkar vote. Yet moving from the first to second election in a relatively high rent district (17.5) correlates with a 10% loss in Golkar support. Golkar’s decline was clearly sharpest in the areas where it lost control of large state bureaucracies. 223 Figure 30 – Golkar’s decline The result is intuitive: Golkar was most likely to decline in those areas where it was previously strong, and the areas it was previously strong were areas with large bureaucracies. The result is robust to a number of alternative specifications strengthening confidence that civil service size is the crucial factor.251 O’Rourke, for instance, has suggested that Golkar succeeded in areas of high poverty. Adding a poverty variable does not disturb the result.252 Also, it is possible that the over-whelming Golkar support in some areas tended to track the existence of desirable leadership. In this logic, Iramasuka voters want to vote for an Iramasuka candidate. However, even adding a dummy for ‘Iramasuka’ leadership does not significantly alter the result.253 Golkar’s decline in high rent areas occurred regardless of who was at the helm. 251 For robustness checks, see Appendix H, Section 5.. 252 This is a simplification of the author’s argument. O’Rourke provides an impressive list of potential causes for Golkar’s success in some areas in 1999, including fraud (209), past policy performance (243), poverty (209), geography (209, 242), ignorance / lack of sophistication (209, 242-3), support for Habibie (217), and last but not least, the control of sub-national resources (209). In the end, though, O’Rourke still finds Golkar’s performance “mystifyingly strong” (O’Rourke 2002, 243). 253 For ‘Iramssuka’ provinces, I coded the variable ‘1’ for 1999 (Sulawesi-born Habibie led the party), ‘0’ in 2004 (North Sumatra-born Tandjung led the party), and ‘1’ for 2009 (Sulawesi-born Kalla led the party). 224 These tests provide support for Hypothesis 3. Golkar support strongly correlated with civil service size in 1999, but the strength and significance of the relationship declined thereafter. Conclusion Choi’s recent explanation of party system expansion underlined the role of executive elections and the rise of the anti-corruption parties. These factors contributed to expansion but they do not explain why party systems fragmented unevenly across the country. This chapter provides an account of party system change within electoral districts. I argued that party system expansion can be explained by evolving expectations of rent sharing. In 1999, voters and elites expected access to state resources would be dominated by one party. As a result, those areas with high rent opportunities tended to vote for large parties, consolidating local systems in the process. As an expectation of rent sharing emerged, so too did a new pattern of partisan organization. Anticipating consensual politics, elites and voters eschewed major parties in favour of minor party options. The shift from centralized machine politics to decentralized clientelist politics fragmented the party system. The argument ties the expansion of the party system directly to the decline of Golkar. Golkar’s dominance in high rent provinces effectively blocked the causal path that would lead from ethnic diversity to electoral fragmentation in 1999. It took several years for elites and voters in high patronage provinces to adjust to the more open political environment of the reformasi era. What replaced Golkar in these regions was not a new monolith but a large number of minor players. Only after the decline of Golkar do we see the fragmentation of the electoral vote in these diverse, high-patronage districts. It should be noted, however, that it was no coincidence Golkar was led into an election by an ‘Iramasuka’ leader two of the three post-Suharto elections. 225 Chapter 9 - Conclusion Research findings Institutional designers in Indonesia have taken aggressive measures to prevent the formation of regional and ethnic parties while simultaneously attempting to reduce the number of partisan competitors. Despite these efforts, this dissertation provides strong evidence that the correlation between ethnic diversity and the effective number of electoral parties is robust and consistent across time. Even going back to the 1999 election, a strong relationship between diversity and party system size existed at the municipal level. Since 1999, the national level electoral districts that have experienced the greatest fragmentation also tend to be the most diverse. This stylized fact, which motivates the dissertation, is a novel finding in its own right. Previous investigations of the Indonesian party system either ignore the issue of district-level diversity or assert that ethnic diversity cannot explain party system change (Choi 2010; D. Y. King 2003). At the very least, this dissertation has made a modest contribution by simply highlighting the importance of the ethnic variable in explaining party system outcomes. While this knowledge itself is useful, I have sought to go beyond simply correlative findings to uncover deeper mechanisms. Why does party system size across electoral districts correlate with ethnic diversity? The correlation is seemingly consistent with a body of cross-national literature that finds a similar correlation between ethnic diversity and party system fragmentation. Still, the established literature has not presented a fleshed out theoretical mechanism that leads from diversity to party system fragmentation. The role of communal voting and ethnicized parties is simply assumed. In Indonesia, the vagueness of the mechanism is particularly problematic because a fundamental link in the causal chain – ethnic parties – does not exist. How, then, does ethnic diversity in Indonesia impact partisan competition in the absence of ethnic parties? This dissertation finds that a simple communal voting story, in which ethnic groups vote for ethnic parties, cannot adequately explain electoral dynamics in Indonesia. Most of the tell-tale signs of ethnic voting are absent. For instance, polling reveals that voters in diverse areas are less likely to hold ethnically chauvinistic political opinions. Likewise, results from the non-partisan upper-house (DPD) elections indicate that diverse 226 electoral districts do not produce a fragmented electoral vote in the absence of political parties. Furthermore, an examination of party system size at the municipal level did not uncover clear evidence of ethnically driven ‘block voting.’ Voters may vote for co- ethnics candidates in large numbers, but this fact by itself does not appear to fragment the party system. To account for the correlation between diversity and electoral fragmentation, I have highlighted the formative role of rent opportunities. Ethnic diversity has an indirect effect on the party system because it shapes state structures and patterns of corruption. In diverse electoral districts, the livelihoods of voters and elites are tightly linked to the control of the state sector. Elites have more opportunities to engage in rent-seeking behavior, which affects the way they participate in the political sphere. First, the opportunity to manipulate state resources draws elites into the electoral arena, increasing the number of viable candidates. Second, the intense focus on local goods distribution diminishes the value of national party platforms, allowing candidates to pursue political office under minor party labels. Third, voter demands for particularistic goods distribution lead them to disregard party labels and form tight patron-client linkages with candidates. The upshot of all these phenomena is the expansion of the electoral vote attained by minor parties, which act as vehicles of convenience for locally oriented rent- seeking networks. In high diversity / high rent electoral districts, the expansion of the vote attained by the minor parties fragments the party system. To establish the plausibility of the argument, the dissertation carefully analyzed distinct pieces of the causal mechanism. Several findings are worth highlighting. First, increased numbers of elites tend to enter electoral politics in poorly governed areas. Controlling for other important variables, those electoral districts with poor infrastructure provision – a proxy for rent opportunities - tended to also have more candidates competing for office. This finding held true even when looking solely at ethnically homogenous electoral districts. The decision to enter national politics in high rent opportunity electoral districts appears to be closely connected efforts to the capture sub- national power. Strong support for this claim comes from the comparison of national and sub-national entry rates. Elites only tend to pursue minor party candidacies when a strong network of sub-national co-partisans exists. The intense competition for sub-national 227 power in high rent areas percolates up to the national level, increasing the number of viable political contenders within an electoral district. Second, candidates in high rent areas were more likely to switch parties. Party career paths provide one measurement of elite investment in party labels. Between 2004 and 2009, a high proportion of candidates switching parties positively correlated with provincial civil service size, a key measure of the rent opportunities concept. The fact that candidates in high rent areas were less likely to stay within the same party indicates an opportunistic disposition toward party labels in certain areas. This finding dovetails with my argument that viable elites in high rent districts are more likely to join minor party labels. When candidates feel they can rely on their personal appeal and resources to win votes they are more willing to float freely across the system. Third, preference voting rates for national candidates in 2004 positively correlated with provincial civil service size. This finding supports my argument that electoral competition in high rent areas is less about programmatic preferences than simply personal ties. While gift giving is ubiquitous throughout Indonesia, candidates in high rent areas tend to put particular emphasis on appealing to voters by promising to provide direct support. The high preference voting rates confirm the tightness of the patron-client bond in these areas. This finding is particularly strong in small parties, which are especially reliant upon the particularistic appeals of local candidates. Fourth, provincial civil service size positively correlated with electoral fragmentation in municipal electoral districts. This finding was strong even in ethnically homogenous municipalities. In high rent areas, there is a strong tendency for voters to support minor electoral parties, even at the national level. Support for minor parties across levels is indicative of locally oriented networks that populate minor parties at multiple levels of governance. Minor party breakthroughs add additional players to the party system, expanding its size. Fifth, provincial civil size negatively correlated with municipal electoral fragmentation in 1999. In the country’s first election, high rent opportunity areas were dominated by Golkar, Suharto’s state-party. The negative correlation underlines the important role rent sharing expectations play in determining the direction of the relationship between rent opportunities and party system size. Many elites and voters 228 supported the Golkar machine in 1999 because they continued to believe the party would dominate the distribution of state favours. After several years of experience with near universal coalitions, in which a broad coalition of parties shared state resources, public expectations of rent sharing evolved. Support for the machine was no longer necessary to access state spoils. Accordingly, between 1999 and 2009, the Golkar vote in high rent areas collapsed. In its place came not another machine but the proliferation of small actors. The dissertation, then, links the fragmentation of the Indonesian party system with the structure of the Indonesian state and patterns of corruption. Contemporary Indonesian politics revolves around the fight for the spoils of office: the power to pass out state jobs, rig contracts, siphon off transfer payments, and sell political support. The relative importance of local political power varies however. Those areas that remain heavily dependent upon state spending produce a distinct style of politics. It is personal rather than programmatic and local rather than national. In these high rent areas, minor party labels find themselves effectively taken over by loose networks of locally oriented rent seekers. These minor parties compete with, and sometimes beat, the major parties that dominate the low rent areas. Due to broad coalitions, these nationally insignificant parties find themselves holding lucrative sub-national seats. My dissertation shows that the potential rewards of office impact the way politicians interact with both the party system and voters. In so doing, I add to a scattered body of Indonesian case literature concerned with the interaction between rents and parties. For example, Slater finds that the need to protect the spoils system has motivated the formation of collusive coalitions (Slater 2004). The early link between Golkar support and civil service size was highlighted by van Klinken (van Klinken 2007). Ufen traces the personalization of the party system back to, among other things, the localization of political competition (Ufen 2008). Choi, too, proposes a mechanism linking corruption and party system change (Choi 2010). I have utilized these insights to provide a broad account of party system size and party system change in Indonesia at the district level. This interpretation of the Indonesian party system departs with previous studies that have focused on broad national level socio-cultural differences between the parties. King’s examination of the 1999 party system emphasized the continuity of aliran based 229 voting patterns from 1955 to the contemporary period (D. Y. King 2003). While the power of the aliran based interpretation has faded, work by Tanuwidjaja reaffirms the commonly held view that Indonesian party politics is driven by religious orientations (Tanuwidjaja 2010). Similarly, Suryadinata et al. forward the Javanese/non-Javanese divide as a formative factor shaping the party system (Suryadinata, Ananta, and Arifin 2004). These works all make important contributions, yet they fail to capture the widespread intuition that contemporary Indonesian political competition is motivated, in large part, by the potential to enjoy the material benefits of office holding. Indonesian politicians invest large sums of money into their careers and expect to reap the rewards once in office. I do not deny that religion plays an important role in Indonesian political life; indeed, several findings from the empirical chapters reaffirm the importance of religion. Still, my study underlines the need to take into account the systemic effects that state spoils have on the party system. Theoretical contributions This study makes four contributions to the literature on comparative party systems. First, my findings challenge the (too often) unstated assumption that communal voting in diverse societies fragments the party system. Ethnic diversity has proven itself to be a productive independent variable but, with the exceptions of Stoll and Mozaffer et al, party systems analysts have failed to grapple with the theoretical implications of the measure itself (Mozaffar, Scarritt, and Galaich 2003; Stoll 2007). None of the canonical works empirically walk through the mechanism that proceeds from group preferences to partisan support to party system outcomes. Indonesia presents a good news / bad news dilemma for existing theory. The good news is that the key independent variable – ethnic diversity – does in fact correlate with the effective number of electoral parties. The bad news is that only weak signs of communal voting exist. Much of the relationship between diversity and party system size is driven by a distinct mechanism that has little to do with communal voting. This is not to say that that communal voting assumptions lack validity. Clearly, some countries exhibit the signs of the ‘census vote’ in which the electorate selects only parties that emerge from their own community. The evolution of the Belgian party 230 system, with its distinct Flemish and Walloon partisan options, lends credence to the underlying intuition. Indeed, studies of Belgium have found an interaction between institutions and ethnicity at the sub-national level (Geys 2006). Nonetheless, one lesson from the Indonesian case is that the existence of the relationship is conditional upon the existence of ethnic options. This finding overlaps with the work of Madrid, who demonstrates that sub-national ethnic diversity in Latin American country does not correlate with fragmented party systems in the absence of ethnic options (Madrid 2005). Sensitivity to the existence of ethnic options needs to be integrated into the broader literature. In highlighting the formative role that internal state structure plays in determining party system size, I make a second contribution to the literature on party system nationalization. This project owes an intellectual debt to those prior works that focused on the interaction between the centralization of authority and candidate decision making (Chhibber and Kollman 2004; Hicken 2009). In Indonesia, decentralization has only strengthened the localization of political competition. The size of the national prize has shrunk with the transfer of fiscal capacities to the sub-national governments. Accordingly, political strategies have evolved. Many elites and voters have chosen to eschew the large national labels in part because national power is a secondary goal to attaining sub-national spoils. This is consistent with the lessons of the party nationalization literature. Where I move the literature forward is to focus on the impact of local resources on district level outcomes. Existing work tends to assume that electoral institutions determine district level party system size. The Indonesian case, however, demonstrates that the struggle for sub-national resources can have district level effects. The provinces and municipalities that receive fewer transfers have relatively smaller sub-national governments. Voters in these areas are more likely to support the large national parties in part because they are less reliant on state largesse and thus can afford to vote for their favourite national leader or preferred policy oriented party. Their focus on the national competition has had a consolidating effect on district level electoral fragmentation, at least in the last two elections. Therefore, where previous literature treats the decentralization of authority as a critical independent variable that shapes party system 231 size in a uniform way across the country, I discover that the effect of the variable is mediated by variance in the size and the relative importance of the sub-national units themselves. I make a third contribution to the literature by treating rent opportunities as an independent variable determining the size of the party system. Scholars have linked state access to the type of parties that emerge in a system and the mode of party-voter interaction that predominates. Shefter ties the dominance and durability of a party machine to the ‘supply of patronage’, Kopecky & Scherlis find that parties that rally supporters using clientelistic exchange typically have a broad ‘scope of party patronage’, and Chandra theorizes that the emergence of ethnic parties is facilitated by the existence of a ‘patronage democracy’ (Chandra 2004; Kopecký and Scherlis 2008; Shefter 1977). In these studies, the role of the state and the prevailing constraints on behavior affect the constituent units within the party system. This dissertation affirms that state resources shape party systems, and moves the literature forward by turning attention to the number of parties. Use of state resources for political advantage has frequently been associated with the activity of impersonal party machines. Walking the logic of the machine forward, we might assume the prevalence of state resources produces a consolidated party system, one dominated by a machine monopolizing valuable state favours. I find that the importance of spoils politics fragments the party system. To explain the counter intuitive finding I have highlighted the important role played by expectations of rent sharing. The dominance of large machines relies on the belief that they will monopolize resources; when they cannot credibly commit to doing so, an abundance of state resources erodes the electoral strength of the potential machine rather than strengthening it. These processes tend to result in an inchoate party system with a myriad of small actors. I make a fourth contribution to the literature on ethic party bans. Only recently has the phenomenon of ethnic party bans made it onto the radar of political scientists. Ethnic bans have been categorized, the causal factors motivating the bans have been explored, and their implementation has been analyzed in various settings (Bogaards 2010; Hartmann and Kemmerzell 2010; Hartmann 2010; Moroff 2010). Work on the effect of bans has focused on variables capturing the incidence of internal violence (Basedau and 232 Moroff 2011). Most of the work has concentrated on African countries. This study adds to the literature by looking closely at the effects of diversity has on the party system once a de-facto ban has been put in place. The evidence suggests the ban is ‘working’ in that political aspirations are channeled into broadly national parties. Despite the suggestive correlation between party system size and ethnic diversity, there is minimal evidence that the current party system merely provides a façade for localized competition among various suku bangsa. Extensions and future research There are several potential extensions of the research agenda set forth in this dissertation. First, the questions I have explored in the Indonesian context should be investigated cross-nationally, beginning with the issue of party system size and communal voting. We know that ethnic diversity correlates with national level party system size, but we have only minimal evidence that the finding is in fact driven by distinct patterns of ethnic voting. Furthermore, the issue of ethnic party bans has not yet been integrated into the literature. Future research should explore the effect of party bans on party system size. Outside of the Indonesian context, do ethnic party bans short-circuit the connection between diversity and party system fragmentation? Do they constrain the practice of communal voting? Even in the absence of ethnic party bans, is the correlation between ethnic diversity and party system size in fact driven by communal voting or some other process altogether? This dissertation found a distinct mechanism connecting fragmentation and diversity that did not rely on ethnic parties and communal voting. Cross-national testing could provide an opportunity to apply the rent opportunities concept beyond Indonesian borders. Rent opportunities can be operationalized through the combination of existing datasets on corruption perceptions and state size. How the rents measure affects electoral and party outcomes will depend on the preexisting structure of the political system. Thus, a cross-national study of rent opportunities and party systems should interact the rents variable with existing variables that measure the centralization of political authority. For instance, we may find that it is the combination of decentralized authority and high rent opportunities that fragment the national party system, while the combination of 233 centralized authority and high rent opportunities consolidates the system. Rent opportunities could end up being an important variable in the study of party system size, though the Indonesian case demonstrates that research needs to be sensitive to conditional effects. Second, the research agenda should be extended through the collection of district level electoral results and demographic information across countries. This study of Indonesia is one of only a few investigations of party systems and ethnicity to consider district level dynamics. The party systems literature tends to establish relationships at the national level but too rarely tests the arguments at the district level. This is unfortunate, considering that mechanisms involving electoral systems and communal voting are typically theorized at the district level. To push the literature forward, researchers need to begin compiling district level ethnicity data across countries. Country studies have collected district level data in Belgium, Latin America, Switzerland, America, and now Indonesia. These efforts need to be aggregated into one dataset so that the task of future collection can be coordinated. Third, we need more research on voter and elite behaviour in multi-level contexts. The argument in this dissertation hinges on the assumption that sub-national political competition percolates up to, and has consequences on, national level outcomes. There is good reason to believe the national and sub-national dynamics are tightly inter-twined: Indonesia voters almost always vote for the same party at multiple levels, candidates only tend to run nationally when there are plenty of co-partisans running for sub-national office under the same label, and sub-national rent opportunities clearly affect a voter’s likelihood of casting a preference vote and/or backing a minor party. Nonetheless, I have only scratched the surface of questions that may be explored in this area. Future research should look at the interaction between co-partisans at different levels between elections. I have hypothesized that national level candidates financially benefit when their co- partisans control sub-national offices. Data from candidate wealth statistics or contract practices may be able to confirm or reject the hypothesis. Likewise, we also need to look at the effect of sub-national office holding on national level electoral results, and vice versa. Can sub-national office holders use their resources to aid national level co- partisans? We might start by looking for a demonstrable vote share increase for 234 candidates who have a co-partisan seated in the local executive. Given the large number of sub-national units in Indonesia, a regression discontinuity research design that utilizes partisan office holding at the municipal level and national level legislative vote share may reveal the existence (or non existence) of a distinct effect of sub-national office holding. Finally, more research is required on the logic of Indonesian coalitions, especially at the sub-national level. This dissertation underlines the importance of oversized Indonesian coalition while only briefly describing their regularity and origin. What is the typical number of parties in a post-election coalition? How are key positions allocated? Are there any ideological underpinnings to coalitions? Is there any relevant variation across the country? Indonesia’s propensity for oversized coalitions has intrigued observers since the 1950s. Given the wealth of post-Suharto sub-national data, investigating patterns in more detail would undoubtedly prove to be a productive line of research. Discussion of policy implications Theoretically, the dissertation is concerned with the causes of party system size. Practically, the dissertation is interested in the effectiveness of Indonesia’s electoral institutions. Are the institutions meeting the goals they were designed to accomplish? Is the party system responding to institutional changes in the ways we expect? Can we foresee any problems on the horizon? To conclude the dissertation, I return to the questions of institutional design that directly relate to the study. Are Indonesia’s strict regional party requirements justified? Modern representative democracy is defined by the right to start parties and challenge the government in a moderately fair election. When the state severely restricts the right to launch parties, as it does in Indonesia, there needs to be just cause. The restrictive laws should effectively contain a viable threat. Indonesia’s wave of transitional violence justified decisive action. The tool used to accomplish the goal – regional requirements on party registration – works as well as anyone could hope. Though parties have areas of regional strength and weakness, they all effectively compete across the country. Outside of Aceh, there are no clear danger areas where secessionist sentiments overlap with the overwhelming presence of one party. In other words, latent regional grievances are not 235 ‘particised’. What’s more, there are only weak signs that local partisan conflicts have been ‘ethnicised’. The dominance of national parties does not simply mask a troubling pattern of sub-national ethnic conflict. And no major societal group has denounced the nationalized party system as unfair and illegitimate. Rather, the evidence we have suggests political aspirations are being effectively channeled through the national parties. This is an impressive feat in such a diverse country that only recently experienced sharp communal violence. My assessment of party system consolidation efforts is less positive. In their efforts to bring the number of parties down, institutional engineers have graduated from mild re-registration requirements to blunt legislative thresholds. At the time of writing, the parties are discussing raising the threshold to 5% of the national electoral vote. No doubt, the threshold of 2009 did help reduce the raw number of legislative parties from 17 to 9. The reduction in the effective number of legislative parties was nowhere near as dramatic, dropping from 7.1 to 6.1. This legislative consolidation occurred in the context of increased electoral fragmentation, from 8.6 to 9.6 parties. Thus the minor consolidation of the legislative system came with a heavy cost in overall proportionality of representation. Will institutional changes influence future political behaviour in ways that consolidate the system? There has only been one election with the legislative threshold in place, so definitive answers remain elusive. In the future, elites may think twice before launching an upstart party. A reduction in launched parties could end up bringing down the effective number of electoral parties, thereby reducing the high number of ‘wasted’ votes. In this optimistic account, the disproportionality of the electoral system will fade out as elites and voters adjust to the new incentives. I do not foresee this occurring. Since 1999, each election has brought new restrictions on minor parties, yet they keep successfully competing for votes. My research indicates that elites and voters may not respond to national level incentives because national level power is not always the prize that captures people’s attention. If national power was the only prize worth fighting for, we might expect national level laws that limit minor party prospects to eventually prevent the launch of long shot partisan contenders. This is not the case in Indonesia, where parties hold thousands of valuable 236 sub-national seats. In a country where politics is a lucrative business, these organizations are unlikely to pack up shop simply because they are denied a few national level offices. The likely consequence of persistent minor party electoral survival is disproportionality, especially within national level electoral districts. The problem is compounded by barring minor parties from holding any seats in the national legislature. In 2004, before the legislative threshold was enacted, only 4.8% of the electoral vote went to parties that received no representation in the national legislature. In 2009, that number jumped to 18.3 %. This group of nationally unrepresented voters is only slightly lower than the group of voters that backed PD, the largest party in the legislature. Clearly, the threshold has increased the number of nationally unrepresented voters. Proponents claim the legislative threshold improves the efficiency of the legislature by reducing the number of players. Though we do not have systematic proof of the claim, they probably have a point. Indonesia’s consensual pattern of legislative life gives undue power to small parties. Nonetheless, the current size of unrepresented voters is abnormally high for a proportional system. Optimistically, voters may be fine without national partisan representation. There is, however, a danger that large scale under- representation could breed disaffection and anti-system attitudes. Ultimately, Indonesia’s institutional framework is trying to accomplish too many tasks. Efforts to nationalize the party system have been effective. But the simultaneous effort to provide relatively proportional representation and legislative efficiency has institutions working at cross purposes. Electoral fragmentation, spurred on by decentralization and direct presidential elections, will continue to place stress on attempts to improve efficiency and representation. Institutional designers should resign themselves to this phenomenon and focus on making the proportional system produce more representative legislative outcomes. Improvements in legislative efficiency should be accomplished through a careful modification of legislative rules and procedures. Concerned institutional designers might start their reform process through experimental rule changes in sub-national legislatures to see which tweaks produce the desired outcomes. The overall strategy, however, should not be a futile attack on party fragmentation, but rather measured reforms to manage the consequences of a challenging but uncontrollable outcome. 237 Bibliography Aldrich, John H. 1995. Why Parties? The Origins and Transformation of Party Politics in America. Chicago: University of Chicago Press. Alesina, Alberto, Reza Baqir, and William Easterly. 1999. “Public goods and ethnic divisions.” The Quarterly Journal of Economics 114(4): 1243-1284. Alesina, Alberto, and Eliana La Ferrara. 2004. “Ethnic diversity and economic performance.” Journal of Economic Literature XLIII(September): 762-800. Anderson, Benedict. 1991. Imagined Communities. 2nd ed. London: Verso. ———. 1983. “Old State , New Society : Indonesia’s New Order in Comparative Historical Perspective.” Journal of Asian Studies 42(3): 477-496. Anderson, Lisa. 1987. “The State in the Middle East and North Africa.” Comparative Politics 20(1): 1-18. Aragon, Lorraine V. 2007. “Elite Competition in Central Sulawesi.” In Renogotiating Boundaries, eds. Gerry van Klinken and Henk Schulte Nordholt. Leiden: KITLV Press, p. 39-66. Aspinall, Edward. 2006. “Violence and Identity Formation in Aceh under Indonesian Rule.” In Verandah of Violence: The Background to the Aceh Problem, Singapore: Singapore University Press, p. 149-176. Badan Pusat Statistik. 2008. Data Dan Informasi Kemiskinan Tahun 2007 - Buku 2. Jakarta: Badan Pusat Statiskik. Baldwin, Kate, and John D. Huber. 2010. “Economic versus Cultural Differences: Forms of Ethnic Diversity and Public Goods Provision.” American Political Science Review 104(04): 644-662. Bandiera, Oriana, and Gilat Levy. 2011. “Diversity and the power of the elites in democratic societies: Evidence from Indonesia.” Journal of Public Economics 95(11-12): 1322-1330. Barter, Shane Joshua. 2011. “The Free Aceh Elections? The 2009 Legislative Contests in Aceh.” Indonesia 91(April): 113-130. Basedau, Matthias, Matthijs Bogaards, and Christof Hartmann. 2007. “Ethnic Party Bans In Africa : A Research Agenda.” German Law Journak 8(6): 617-634. Basedau, Matthias, and Anika Moroff. 2011. “Parties in chains: Do ethnic party bans in Africa promote peace?” Party Politics 17(2): 205-222. 238 Bertrand, Jacques. 2004. Nationalism and Ethnic Conflict in Indonesia. Cambridge: Cambridge University Press. Besley, T. 1997. “An economic model of representative democracy.” The Quarterly Journal of Economics (February): 85-114. van Biezen, I., and P. Kopecky. 2007. “The State and the Parties: Public Funding, Public Regulation and Rent-Seeking in Contemporary Democracies.” Party Politics 13(2): 235-254. Blais, Andre, and R. K. Carty. 1990. “Does proportional representation foster voter turnout ?” European Journal of Political Research 18(1990): 167-181. Blais, Andre, and Angieszka Dobrzynska. 1998. “Turnout in electoral democracies ´.” European Journal of Political Research 33: 239-261. Bogaards, Matthijs. 2010. “Ethnic party bans and institutional engineering in Nigeria.” Democratization 17(4): 730-749. Brambor, Thomas, William Roberts Clark, and Matt Golder. 2007. “Are African party systems different?” Electoral Studies 26(2): 315-323. ———. 2005. “Understanding Interaction Models: Improving Empirical Analyses.” Political Analysis 14(1): 63-82. Brown, Colin. 2003. Indonesia: The Unlikely Nation? Crows Nest NSW: Allen & Unwin. Brown, Graham K. 2005. “Playing the (Non)Ethnic Card : The Electoral System and Ethnic Voting Patterns in Malaysia.” Ethnopolitics 4(4): 439-445. Buchanan, James M. 1980. “Rent Seeking and Profit Seeking.” In Toward a Theory of the Rent-Seeking Society, eds. James M Buchanan, Robert D Tollison, and Gordon Tullock. College Station: Texas A&M University Press, p. 3-15. Budge, Ian, and Michael D. McDonald. 2007. “Election and party system effects on policy representation: Bringing time into a comparative perspective.” Electoral Studies 26(1): 168-179. Budiardo, Miriam. 2001. “The Electoral System (Points for Discussion).” In ed. R. William Liddle. Bandung: Penerbit Mizan, p. 127-131. Buehler, Michael, and Paige Johnson Tan. 2007. “Party-Candidate Relationships in Indonesian Local Politics: A Case Study of the 2005 Regional Elections in Gowa, South Sulawesi Province.” Indonesia 84: 41-69. 239 Cain, Bruce E, John A. Ferejohn, and Morris P. Fiorina. 1987. The Personal Vote: Constituency Service and Electoral Independence. Cambridge: Harvard University Press. Carey, John M, and Matthew Soberg Shugart. 1995. “Incentives to cultivate a personal vote: A rank ordering of electoral formulas.” Electoral studies 14(4): 417-439. Carson, Jamie L. 2005. “Strategy , Selection , and Candidate Competition in U.S. House and Senate Elections.” Journal of Politics 67(1): 1-28. Chandra, Kanchan. 2007. “Counting Heads: A Theory of Voter and Elite Behaviour in Patronage-Democracies.” In Patrons, Clients and Policies: Patterns of Democratic Accountability and Political Competition, eds. Herbert Kitschelt and Steven Wilkinson. Cambridge: (Cambridge University Press. ———. 2004. Why Ethnic Parties Succeed. Cambridg: Cambridge University Press. Chandra, Kanchan, and Steven Wilkinson. 2008. “Measuring the Effect of ‘Ethnicity’.” Comparative Political Studies 41(4-5): 515-563. Chhibber, Pradeep, and Ken Kollman. 1998. “Party Aggregation and the Number of Parties in India and the United States.” American Political Science Review 92(2): 329-342. ———. 2004. The Formation of National Party Systems: Federalism and Party Competition in Canada, Great Britain, India, and the United States. Princeton: Princeton University Press. Chhibber, Pradeep, and Irfan Nooruddin. 2004. “Do Party Systems Count?: The Number of Parties and Government Performance in the Indian States.” Comparative Political Studies 37(2): 152-187. Choi, Jungug. 2010. “District Magnitude , Social Diversity , and Indonesia’s Parliamentary Party System from 1999 to 2009.” Asian Survey 50(4): 663-683. Clark, Terry D, and Jill N Wittrock. 2005. “Presidentialism and the Effect of Electoral Law in Postcommunist Systems: Regime Type Matters.” Comparative Political Studies 38(2): 171-188. Clark, William Roberts, and Matt Golder. 2006. “Rehabilitating Duverger’s Theory: Testing the Mechanical and Strategic Modifying Effects of Electoral Laws.” Comparative Political Studies 39(6): 679-708. Collier, Paul. 2000. “Ethnicity, politics and economic performance.” Economics and Politics 12(3). 240 Cox, Gary. 1997. Making Votes Count: Strategic Coordination in the World’s Electoral Systems. Cambridge: Cambridge University Press. Cox, Gary, and Octavio Amorim Neto. 1997. “Electoral institutions, cleavage structures, and the number of parties.” American Journal of Political Science 41(1): 149-174. Cox, Gary, and Frances Rosenbluth. 1994. “Reducing nomination errors: Factional competition and party strategy in Japan.” Electoral Studies 13(1): 4-16. Crouch, Harold. 2010. Political Reform in Indonesia After Sueharto. Singapore: ISEAS. ———. 1988. The Army and Politics in Indonesia. 2nd ed. Ithaca: Cornell University Press. Davidson, Jamie S. 2007. “Politics-as-usual on trial: regional anti-corruption campaigns in Indonesia.” The Pacific Review 20(1): 75-99. Desposato, Scott W. 2009. “Parties for Rent? Ambition, Ideology, and Party Switching in Brazil’s Chamber of Deputies.” American Journal of Political Science 50(1): 62-80. Easterly, William, and Ross Levine. 1997. “Africa’s growth tragedy: policies and ethnic divisions.” The Quarterly Journal of Economics 112(4): 1203-1250. Elson, R.E. 2001. Suharto: A Political Biography. Cambridge: Cambridge University Press. Evans, Kevin Raymond. 2003. The History of Political Parties & General Elections in Indonesia. Jakarta: Arise Consultancies. Feddersen, Timothy J., Itai Sened, and Stephen G. Wright. 1990. “Rational Voting and Candidate Entry under Plurality Rule.” American Journal of Political Science 34(4): 1005-1016. Feith, Herbert. 1962. The Decline of Constitutional Democracy in Indonesia. Ithaca: Cornell University Press. ———. 1957. The Indonesian elections of 1955. Ithaca: Cornell University Press. Feith, Herbert, and Lance Castles. 1970. Indonesian Political Thinking 1945-1965. Ithaca: Cornell University Press. Ferree, Karen E. 2006. “Explaining South Africa’s Racial Census.” Journal of Politics 68(4): 803-815. 241 Filippov, Mikhail G, Peter C Ordeshook, and Olga V Shvetsova. 1999. “Party Fragmentation and Presidential Elections in Post-Communist Democracies.” Constitutional Political Economy 10: 3-26. Fitrani, Fitria, Bert Hofman, and Kai Kaiser. 2005. “Unity in diversity? The creation of new local governments in a decentralising Indonesia.” Bulletin of Indonesian Economic Studies 41(1): 57-79. Geys, Benny. 2006. “District Magnitude, Social Heterogeneity and Local Party System Fragmentation.” Party Politics 12(2): 281-297. Glaeser, Edward L., and Raven E. Saks. 2006. “Corruption in America.” Journal of Public Economics 90(6-7): 1053-1072. Golder, Matt. 2006. “Presidential Coattails and Legislative Fragmentation.” American Journal of Political Science 50(1): 34-48. Habyarimana, James et al. 2007. “Why does ethnic diversity undermine public goods provision?” American Political Science Review 101(4): 709-725. Hadar, Leon T. 2000. 1-19 Averting a “New Kosovo” inIndonesia: Opportunities and Pitfalls for the United States. Washington. Hadiz, Vedi. R. 2003. “Power and Politics in North Sumatra: The Uncompleted Reformasi.” In Local Power and Politics in Indonesia: Decentralisation & Democratisation, eds. Edward Aspinall and Greg Fealy. Singapore: ISEAS, p. 119- 131. Hartmann, Christof. 2010. “Senegal’s party system: the limits of formal regulation.” Democratization 17(4): 769-786. Hartmann, Christof, and Jörg Kemmerzell. 2010. “Understanding variations in party bans in Africa.” Democratization 17(4): 642-665. Heller, William B, and Carol Mershon. 2005. “Party Switching in the Italian Chamber of Deputies , 1996 – 2001.” Journal of Politics 67(2): 536-559. Henley, David. 1995. “Ethnogeographic Integration and Exclusion in Anticolonial Nationalism : Indonesia and Indochina.” Comparative Studies in Society and History 37(2): 286-324. Hicken, Allen. 2009. Building Party Systems in Developing Democracies. New York: Cambridge University Press. ———. 2011. “Clientelism.” Annual Review of Political Science 14(1): 289-310. 242 Hicken, Allen, and Heather Stoll. 2008. “Electoral Rules and the Size of the Prize: How Political Institutions Shape Presidential Party Systems.” The Journal of Politics 70(04): 1109. Holmsten, Stephanie S., Robert G. Moser, and Mary C. Slosar. 2009. “Do Ethnic Parties Exclude Women?” Comparative Political Studies 43(10): 1179-1201. Horowitz, Donald. 2003. “Electoral Systems: A Primer for Decision Makers.” Journal of democracy 14(4): 115-127. ———. 1985. Ethnic groups in conflict. Berkley: University of California Press. ———. 2006. “Strategy Takes a Holiday: Fraenkel and Grofman on the Alternative Vote.” Comparative Political Studies 39(5): 652-662. ———. 2001. “The Draft Laws on Indonesian Political Parties, Elections, and Legislative Bodies: An Analysis.” In Crafting Indonesian Democracy, ed. R. William Liddle. Bandung: Mizan Media Utama, p. 137-147. Indrayana, Denny. 2008. Indonesian Constitutional Reform 1999-2002: An Evaluation of Constitution-Making in Transition. Jakarta: Kompas. International Crisis Group. 2008. 1-16 Indonesia: Pre-Election Anxieties in Aceh. Jackman, Robert W. 1987. “Political Institutions and Voter Turnout in the Industrial Democracies.” American Political Science Review 81(2): 405-424. Jacobson, Gary C. 1989. “Strategic Politicians and the Dynamics of U.S. House Elections, 1946-86.” American Political Science Review 83(3): 773-793. Johnson, Joel W., and Jessica S. Wallack. “Electoral Systems and the Personal Vote.” //hdl.handle.net/1902.1/17901 (Accessed June 8, 2012). Kahin, George McTurnan. 1952. Nationalism and Revolution in Indonesia. Ithaca: Cornell University Press. Kasuya, Yuko. 2008. Presidential Bandwagon: Parties and Party Systems in the Philippines. Manila: Anvil Publishing. Khwaja, Asim Ijaz. 2009. “Can good projects succeed in bad communities?” Journal of Public Economics 93(7-8): 899-916. King, Dwight Y. 2003. Half-Hearted Reform: Electoral Institutions and the Struggle for Democracy in Indonesia. Westport: Praeger. 243 King, Peter. 2008. “‘Corruption Ruins Everything’: Gridlock over Suharto’s Legacy in Indonesia (Part II).” Asia-Pacific Journal: Japan Focus. http://www.japanfocus.org/-Peter-King/2680 (Accessed June 8, 2012). Kitschelt, Herbert. 2000. “Linkages between Citizens and Politicians in Democratic Polities.” Comparative Political Studies 33(6-7): 845-879. van Klinken, Gerry. 2007. Communal violence and democratization in Indonesia: small town wars. London: Routledge. ———. 2003. “Ethnicity in Indonesia.” In Ethnicity in Asia, ed. Colin Mackerras. London: RoutledgeCurzon. van Klinken, Gerry, and Henk Schulte Nordholt. 2007. “Introduction.” In Renogotiating Boundaries, Leiden: KITLV Press, p. 1-29. Komite Pemantuan Pelaksanaan Otonmi Daerah. 2008. Local Economic Governance in Indonesia: A Survey of Businesses in 243 Regencies/Cities in Indonesia, 2007. ———. 2003. Regional Investment Attractiveness: A Survey of Business Perception. Kopecký, Petr, and Gerardo Scherlis. 2008. “Party Patronage in Contemporary Europe.” European Review 16(03): 355-371. Korte, Nina. 2011. “It ’ s Not Only Rents : Explaining the Persistence and Change of Neopatrimonialism in Indonesia.” (167). Krueger, Anne O. 1974. “The Political Economy of the Rent-Seeking Society.” American Economic Review 64(3): 291-303. Kunovich, Sheri, and Pamela Paxton. 2005. “Pathways to Power : The Role of Political Parties in Women ’ s National Political.” American Journal of Sociology 111(2): 505-552. Legge, J.D. 1961. Central Authority and Regional Autonomy in Indonesia: A Study of Local Administration, 1950-1960. Ithaca: Cornell University Press. ———. 2003. Sukarno: A Political Biography. 2nd ed. Singapore: Archipeligo Press. Liddle, R. William. 1970. Ethnicity, Party, and National Integration: An Indonesian Case Study. New Haven: Yale University Press. Liddle, R. William, and S. Mujani. 2007. “Leadership, Party, and Religion: Explaining Voting Behavior in Indonesia.” Comparative Political Studies 40(7): 832-857. 244 Lijphart, Arend. 1991. “Constitutional Choices for New Democracies.” Journal of democracy 2(1): 72-84. ———. 2004. “Constitutional Design for Divided Societies.” Journal of Democracy 15(2): 96-109. Lipset, Seymour Martin, and Stein Rokkan. 1967. “Cleavage Structures, Party Systems, and Voter Alignments: An Introduction.” In Party systems and voter alignments: cross-national perspectives, eds. Seymour Martin Lipset and Stein Rokkan. New York: Free Press. Lipson, Leslie. 1964. The democratic civilization. New York: Oxford University Press. Madrid, Raúl L. 2005. “Indigenous voters and party system fragmentation in Latin America.” Electoral Studies 24(4): 689-707. Mainwaring, Scott. 1993. “Presidentialism, Multipartism, and Democracy: The Difficult Combination.” Comparative Political Studies 26(2): 198-228. Malley, Michael S. 2003. “New Rules, Old Structures and the Limits of Democratic Decentralisation.” In Local Power and Politics in Indonesia: Decentralisation & Democratisation, eds. Edward Aspinall and Greg Fealy. Singapore: ISEAS, p. 102- 116. Marsh, M. 2007. “Candidates or Parties?: Objects of Electoral Choice in Ireland.” Party Politics 13(4): 500-527. Mauro, Paolo. 1995. “Corruption and Growth.” Quarterly Journal of Economics 110(3): 681-712. McIntyre, Angus. 2005. The Indonesian Presidency: The Shift From Personal Toward Constitutional Rule. (Lanham, MD: Rowman & Littlefield. McRae, Kenneth D. 1974. Consociational democracy : political accommodation in segmented societies. Toronto: McClelland and Stewart. Meisel, John. 1974. Cleavages, parties and values in Canada. London: Sage Publications. Mershon, Carol, and Olga V Shvetsova. 2008. “Parliamentary Cycles and Party Switching in Legislatures.” Comparative Political Studies 41(1): 99-127. Mietzner, Marcus. 2009. Military Politics, Islam, and the State in Indonesia: From Turbulent Transition to Democratic Consolidation. Singapore: ISEAS. 245 Miguel, Edward, and Mary Kay Gugerty. 2005. “Ethnic diversity, social sanctions, and public goods in Kenya.” Journal of Public Economics 89(11-12): 2325-2368. Morelli, Massimo. 2004. “Party Formation and Policy Outcomes under Different Electoral Systems.” Review of Economic Studies 71(3): 829-853. Moroff, Anika. 2010. “Party bans in Africa – an empirical overview.” Democratization 17(4): 618-641. Mozaffar, Shaheen, James R. Scarritt, and Glen Galaich. 2003. “Electoral Institutions, Ethnopolitical Cleavages, and Party Systems in Africa’s Emerging Democracies.” American Political Science Review 97(03): 379-390. Nasution, Adnan Buyung. 2001. “Some Notes on Electoral Law Reform in Indonesia.” In Crafting Indonesian Democracy, ed. R. William Liddle. Bandung: Mizan Media Utama, p. 133-136. National Democratic Institute. 1999. The 1999 Presidential Election and Post-Election Developments in Indonesia: A Post-Election Assessment Report. Nichter, Simeon. 2008. “Vote Buying or Turnout Buying? Machine Politics and the Secret Ballot.” American Political Science Review 102(01): 19-31. Norris, Pippa. 1997. Passages to Power: Legislative Recruitment in Advanced Democracies. Cambridge: Cambridge University Press. Ordeshook, Peter C, and Olga V Shvetsova. 1994. “Ethnic Heterogeneity, District Magnitude, and the Number of Parties.” American Journal of Political Science 38(1): 100-123. Osborne, Martin J., and Al Slivinski. 1996. “A model of political competition with citizen-candidates.” The Quarterly Journal of 111(1): 65-96. O’Rourke, Kevin. 2002. Reformasi: The Struggle for Power in Post-Soeharto Indonesia. Crows Nest NSW: Allen & Unwin. Panizza, Ugo. 1999. “On the determinants of fiscal centralization: Theory and evidence.” Journal of Public Economics 74(1): 97-139. Persson, Torsten, and Guido Tabellini. 2000. Political Economics - Explaining Economic Policy. Cambridge: MIT Press. Powell, G. Bingham Jr. 2000. Elections as Instruments of Democracy: Majoritarian and Proportional Visions. New Haven: Yale University Press. 246 Powell, G. Bingham Jr, and Georg S Vanberg. 2000. “Election Laws, Disproportionality and Median Correspondence: Implications for Two Visions of Democracy.” British Journal of Political Science 30(3): 383-411. Przeworski, Adam et al. 1996. “What makes democracies endure?” Journal of democracy 7(1): 39-55. Putnam, Robert D. 2007. “E Pluribus Unum : Diversity and Community in the Twenty- first Century The 2006 Johan Skytte Prize Lecture.” Scandinavian Political Studies 30(2): 137-174. Rabushka, Alvin, and Kenneth A. Shepsle. 1972. Politics in plural societies; a theory of democratic instability. Columbus: Merrill. Reed, Steven R. 2009. “Party Strategy or Candidate Strategy: How Does the LDP Run the Right Number of Candidates in Japan’s Multi-Member Districts?” Party Politics 15(3): 295-314. ———. 1990. “Structure and Behaviour : Extending Duverger ’ s Law to the Japanese Case.” British Journal of Political Science 20(3): 335-356. Reilly, Benjamin. 2006. Democracy and Diversity: Political Engineering in the Asia- Pacific. Oxford: Oxford University Press. ———. 2001. Democracy in Divided Societies: Electoral Engineering for Conflict Management. Cambridge: Cambridge University Press. Remmer, Karen L. 2007. “The Political Economy of Patronage: Expenditure Patterns in the Argentine Provinces, 1983–2003.” Journal of Politics 69(2): 363-377. Reynolds, Andrew. 2002. The Architecture of Democracy: Constitutional Design, Conflict Management, and Democracy. Oxford: Oxford University Press. Riker, William H. 1982. “The Two-party System and Duverger’s Law : An Essay on the History of Political Science.” American Political Science Review 76(4): 753-766. Rose, Richard, and Derek Urwin. 1969. “Social Cohesion, Political Parties and Strains in Regimes.” Comparative Political Studies 2(7): 7-67. Ross, Michael L. 2001. “Does Oil Hinder Democracy?” World Politics 53(3): 325-361. Samuels, David J. 2003. Federalism, and Legislative Politics in Brazil. Cambridge: Cambridge University Press. ———. 1999. “Incentives to Cultivate a Party Vote in Candidate-centric Electoral Systems: Evidence from Brazil.” Comparative Political Studies 32(4): 487-518. 247 ———. 2002. “Presidentialized Parties: The Separation of Powers and Party Organization and Behavior.” Comparative Political Studies 35(4): 461-483. Sartori, Giovanni. 1966. “European Political Parties: The Case of Polarized Pluralism.” In Political Parties and Political Developmen, eds. Joseph LaPalombara and Myron Weiner. Princeton: : Princeton University Press, p. 137-176. Schiller, A. Arthur. 1955. The Formation of Federal Indonesia 1945-1949. The Hague: W. van Hoeve Ltd. Shaffer, Frederic Charles. 2007. “How Effective is Voter Education?” In Elections for Sale: The Causes and Consequences of Vote Buying, ed. Frederic Charles Schaffer. Boulder: Lynne Rienner, p. 161-179. Shefter, Martin. 1977. “Party and Patronage: Germany, England, and Italy.” Politics & Society 7(4): 403-451. Shugart, Matthew Soberg. 2008. “Inherent and Contigent Factors in Reform Initiation in Plurality Systems.” In To Keep or to Change First Past the Post: The Politics of Electoral Reform, ed. Andre Blais. Oxford: Oxford University Press. ———. 1995. “The Electoral Cycle and Institutional Sources of Divided Presidential Government.” American Political Science Review 89(2): 327-343. Shugart, Matthew Soberg, Melody Ellis Valdini, and Kati Suominen. 2005. “and Personal Vote-Earning Attributes of Legislators.” American Journal of Political Science 49(2): 437-449. Shugart, Matthew Soberg, and Martin P. Wattenberg. 2003. Mixed-Member Electoral Systems: The Best of Both Worlds? Oxford: Oxford University Press. Slater, Dan. 2004. “Indonesia’s Accountability Trap: Party Cartels and Presidential Power After Democratic Transition.” Indonesia 78: 61-92. Stokes, Susan C. 2005. “Perverse Accountability : A Formal Model of Machine Politics with Evidence from Argentina.” American Political Science Review 99(3): 315-325. Stoll, Heather. 2007. “Social Cleavages and the Number of Parties: How the Measures You Choose Affect the Answers You Get.” Comparative Political Studies 41(11): 1439-1465. Stone, Walter J., and L. Sandy Maisel. 2003. “The Not-So-Simple Calculus of Winning: Potential U.S. House Candidates’ Nomination and General Election Prospects.” Journal of Politics 65(4): 951-977. 248 Suryadinata, Leo, Aris Ananta, and Evi Nurvidya Arifin. 2004. Indonesian Electoral Behaviour: A Statistical Perspective. Singapore: ISEAS. Taagepera, Rein. 2007. Predicting Party Sizes: The Logic of Simple Electoral Systems. Oxford: Oxford University Press. Taagepera, Rein, and Matthew Soberg Shugart. 1989. Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven: Yale University Press. Tan, Paige Johnson. 2002. “Anti-Party Reaction in Indonesia : Causes and Implications.” Contemporary Southeast Asia 24(3): 484-508. ———. 2006. “Indonesia Seven Years after Soeharto : Party System Institutionalization in a New Democracy.” Contemporary Southeast Asia 28(1): 88-114. Tanasaldy, Taufiq. 2007. “Ethnic Identity Politics in West Kalimantan.” In Renogotiating Boundaries, eds. Gerry van Klinken and Henk Schulte Nordholt. Leiden: KITLV Press, p. 349-371. Tanuwidjaja, Sunny. 2010. “Political Islam and Islamic Parties in Indonesia Critically Assessing the Evidence of Islam’s Political Decline Political Islam and Islamic Parties in Indonesia: Critically Assessing the Evidence of Islam’s Political Decline.” Contemporary Southeast Asia 32(1): 29-49. Tavits, Margit. 2010. “Effect of Local Ties On Electoral Success and Parliamentary Behaviour: The Case of Estonia.” Party Politics 16(2): 215-235. Tomsa, Dirk. 2006. “The Defeat of Centralized Paternalism: Factionalism, Assertive Regional Cadres, and the Long Fall of Golker Chariman Akbar Tandjung.” Indonesia 81(April 2006): 1-22. Transparency International Indonesia. 2010. Mengukur Tingkat Korupsi Kotakota di Indonesia: Denpasar Dipersepsikan Bersih, Pekanbaru dan Cirebon Bermasalah. ———. 2008. Mengukur Tingkat Korupsi di Indonesia: Indeks Persepsi Korupsi Indonesia 2008 dan Indeks Suap. Tsebelis, George. 2002. Veto Players: How Political Institutions Work. Princeton: Princeton University Press. Turner, Mark et al. 2003. Decentralisation in Indonesia: Redesigning the State. Canberra: Asia Pacific Press. USAID. 2006. Stock Taking on Indonesia’s Recent Decentralization Reforms. Jakarta. 249 Ufen, Andreas. 2008. “From aliran to dealignment: political parties in post-Suharto Indonesia.” South East Asia Research 16(1): 5-41. ———. 2006. “Political Parties in Post-Suharto Indonesia: Between Politik Aliran and ‘Philippinisation’.” (December). Valdini, Melody Ellis. 2006. “Electoral Institutions and Information Shortcuts: The Effect of Decisive Intraparty Competition on the Behavior of Voters and Party Elites.” University of California-San Diego. Vatter, Adrian. 2003. “Legislative Party Fragmentation in Swiss Cantons: A Function of Cleavage Structures or Electoral Institutions?” Party Politics 9(4): 445-461. Volden, Craig, and Clifford J. Carrubba. 2004. “The Formation of Oversized Coalitions in Parliamentary Democracies.” American Journal of Political Science 48(3): 521- 537. Waisman, Carlos H, and Arend Lijphart. 1996. Institutional Design in New Democracies: Eastern Europe and Latin America. Boulder: Westview Press. Wanandi, Jusuf. 2002. “Indonesia: A failed state?” The Washington Quarterly 25(3): 135-146. Weingast, Barry R. 1979. “A Rational Choice Perspective on Congressional Norms.” American Journal of Political Science 23(2): 245-262. Ziegenhain, Patrick. 2008. The Indonesian Parliament and Democratization. Singapore: ISEAS. 250 Appendices Appendix A – Map of Indonesia254 Figure 31 - Appendix A – Map of Indonesia 254 Map from taken from Wikimedia Commons database. 251 Appendix B – Supplement to Chapter 2 District level outcomes are, for the most part, not driven by a failure to aggregate across municipal units. While some intra-district party system inflation does exist, it does not correlate with ethnic diversity. This section reinforces that point through multivariate tests on the determinants of party system inflation. I examine party system inflation at two levels: 1) inflation within national electoral districts; 2) inflation within municipalities. Below I describe the variables and statistical tests used at each level. Section 1 - Inflation within national districts Dependent Variable: District-level party system inflation There are two different measures of district-level party system inflation 1. District Inflation (Municipal Total) = Aggregate municipal ENEP – Avg Municipal ENEP 2. District Inflation (National Total) = Aggregate national ENEP - Avg Municipal ENEP Each of the two possible dependent variables has benefits and drawbacks. The benefit of District Inflation (Municipal Total) is that data for both terms are drawn from the same municipal-level dataset. It is not hampered by the issue of split-ticket voting, thus it is the most accurate measure of Party System Inflation. The benefit of District Inflation (National Total) is that it uses data from the key variable of interest in the dissertation: district-level party system size. Because it is drawn from the same data it is a closer measure of national-level dynamics. Rather than decide which is most appropriate, I ran tests on both. The variables were both measured at the geographic level of the national electoral district. Average Municipal ENEP takes the aggregate municipal party system size for all municipalities that lie in the borders of a national electoral district and provides an average score. The electoral results used to calculate these scores came from municipal (DPRDII) vote totals. An example appears below: 252 Table 24 - Appendix B – Party system inflation example Average Municipal ENEP Example National District: North Sumatra 1 Total Kabupaten: 4 Aggregate Municipal ENEP 10.75 Aggregate National ENEP 10.33 Kabupaten Name Aggregate Municipal ENEP Deli Serdang 10.18 Medan 10.19 Serdang Berdagi 9.55 Tebing Tinggi 7.25 Avg Municipal ENEP 9.29 District Inflation (Municipal Total) 1.45 District Inflation (National Total) 1.03 Average Municipal ENEP was subtracted from two different values to arrive at two different inflation scores. First, aggregate municipal ENEP was calculated by adding all the municipal-level votes cast for a party within the geographic confines of a national electoral district. This was the exact same data used to generate the Avg Municipal ENEP. The second value was the district level ENEP calculated using vote totals from the national level. Not all voters cast the exact same vote at all levels. Because there is a slight tendency for voters to abandon small parties at the national level, aggregate national-level ENEP tends to be slightly lower than aggregate municipal-level ENEP. Independent Variables  Civil Service Size – Provincial % of modern sector workers employed in civil service  Ethnic fractionalization – 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous  Urbanization – 0-1 measure, 0 being no 0% of citizens living in areas designated ‘kota’ by central government, 1 being 100% of citizens living in areas designated ‘kota.’  District Magnitude – Average of all Municipal District Magnitude Averages. Logged.  Total Municipalities - Total number of municipalities within the national district 253  Avg Municipal ENEP – the average effective number of electoral parties for all municipalities within the geographic confines of the national electoral district. Results appear below. Models 1 and 2 examine the correlates of district inflation using municipal electoral returns. Models 3 and 4 use national returns. Neither civil service size nor ethnic fractionalization has any statistically significant correlation with party system inflation in any of the models. However, the relationship between civil service size and District Inflation (National Total) does approach statistical significance. Table 25 - Appendix B – Determinants of party system inflation (national) Determinants of Party System Inflation – National Level (OLS Regression) Dependent Variable: Party System Inflation Model 1 - Municipal Vote Totals Model 2 - Municipal Vote Totals Model 3 - National Vote Totals Model 4 - National Vote Totals Civil Service Size -0.37 (2.26) -0.82 (2.09) -3.64 (2.48) -3.96 (2.42) Ethnic Fractionalization 0.54 (0.37) -0.14 (0.40) -0.16 (0.41) -0.65 (0.46) District Magnitude -0.58 (1.29) -1.29 (1.21) -0.27 (1.42) -0.78 (1.40) Total Municipalities 0.07** (0.03) 0.07** (0.03) 0.06* (0.03) 0.05 (0.03) Urbanization 0.10 (0.52) -0.05 (0.48) 0.71 (0.57) 0.60 (0.55) Kab ENEP Avg 0.20*** (0.06) 0.14** (0.07) Const 0.93 (2.74) 1.44 (2.53) 0.23 (3.00) 0.59 (2.93) Obs 67 67 67 67 R2 0.2749 0.3916 0.0995 0.1616 *p < .10. **p < .05. *** p <.01. Section 2 - Inflation within Municipalities Dependent Variable: Municipal Party System Inflation This variable was calculated by subtracting the average effective number of parties within a municipal electoral district by the aggregate effective number of parties within the entire municipality. Municipal Party System Inflation = Aggregate Municipal ENEP - Avg ENEP (municipal district. 254 Independent Variables  Civil Service Size – Provincial % of modern sector workers employed in civil service  Ethnic fractionalization – 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous  City – 1 if municipal unit designated “kota”, 0 if not.  Average District Magnitude (logged) – The average district magnitude of all electoral districts within a municipality.  Total Districts – Total number of electoral districts within the municipality  Average ENEP – The average effective number of parties with a municipal district Two models were run. Model 1 does not include Average ENEP while Model 2 does. Party system inflation is always closely related to Average ENEP. A difficult test for variable is to examine whether it stays consistently signed and significant with and without Average ENEP in the model. Table 26 - Appendix B – Determinants of party system inflation (municipal) Determinants of Party System Inflation – Municipal Level (OLS Regression) Dependent Variable: Party System Inflation Model 1 Model 2 Civil Service Size 0.01** (0.006) 0.01** (0.005) Ethnic Fractionalization -0.02 (0.11) -0.31*** (0.10) District Magnitude -0.50*** (0.18) -0.52*** (0.16) Total Districts -0.05* (0.03) -0.05 (0.03) Kota -0.15* (0.08) -0.24*** (0.07) Average ENEP 0.14*** (0.01) Const 1.61** (0.50) 0.83* (0.45) Obs 432 432 R2 0.0795 0.2773 *p < .10. **p < .05. *** p <.01. Ethnic fractionalization is never positively correlated with inflation. When Average ENEP is added to the model, the negative correlation between ethnic fractionalization and inflation becomes statistically significant. Civil service size is positively correlated with inflation in both models. All else being equal, moving from a moderately low percentage of civil servants (10%) to a moderately 255 high percentage (20%) produces a 0.1 increase in the inflation measure. This is a modest effect, though it is consistent across all models. In sum: there is little evidence of an elite failure to aggregate efforts across electoral districts. Party system inflation both within national electoral districts and within municipalities is very low. The small amount of inflation that does exist does not correlate with ethnic diversity. Figure 32 - Appendix B – Municipal party system inflation 256 Appendix C – Supplement to Chapter 3 Section 1 - Civil service size, ethnic diversity, and population size Figure 5 demonstrates the correlation between ethnic diversity and civil service size. Here I establish the robustness of the relationship across time. As well, I show that a substantial portion of the relationship occurs due to a negative correlation between ethnic diversity and population size. Dependent Variable: Percentage of modern sector workers employed in the civil service This provincial-level variable was constructed by dividing the total number of civil servants by the total number of modern sector workers. A modern sector worker was defined here as a non-agricultural worker. Data was drawn from Indonesian statistical yearbooks. Below I include four snapshots of civil service size: 1990, 1999, 2005, and 2009. The first draws on data from van Klinken (2007) and captures civil service size following the oil boom when the Indonesian bureaucracy was at its largest. The second gives civil service size before the first democratic election. The third and fourth capture civil service size around the times of the 2004 and 2009 elections. Together there are two sets of observations from before the democratic era and two sets of observations after the transition. Independent Variables  Provincial Ethnic Fractionalization - 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous. Measured at the provincial level  Population Size - Total population living within the municipality (logged) Models 1-4 show a clear correlation between ethnic diversity and civil service size is apparent in all four years: the more ethnically diverse a province, the greater the relative proportion of modern sector workers employed by the government. Correlations between these two variables report a relationship significant at the .05 level both before and after the transition to democracy. 257 Table 27 - Appendix C – Determinants of civil service size Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 1990 1999 2005 2009 1990 - Population 1999 - Population 2005 - Population 2009 - Population Estimate Estimate Estimate Estimate Estimate Estimate Estimate Estimate (std. err) (std. err) (std. err) (std. err) (std. err) (std. err) (std. err) (std. err) Population (logged) -5.30*** -1.38 -2.83*** -1.01 -3.13*** -1.01 -2.79*** -0.81 Observations 26 26 30 33 26 26 30 33 R2 0.3485 0.2972 0.2765 0.2739 0.6028 0.4766 0.4671 0.4819 18.822*** (5.25) Ethnic Fractionalization 10.99** -4.66 7.14** -3.27 8.56** -3.51 6.68** -2.99 92.50*** (22.52) 54.70*** (16.34) 49.53*** (13.16) 10.90*** (3.19) 12.34*** (3.77) 10.84*** (3.40) 51.13*** 16.5 const 6.86* (3.79) 5.19** (2.45) 4.44 (2.71) 4.38* (2.32) Proportion of civil servants to non-civil servants (%) Variable 258 The size of the coefficient suggests that relationship is not only statistically significantly but substantively large. The magnitude of the relationship can be demonstrated using simulations. For example, in 2005 the model predicts that a province with an ethnic fractionalization score around the mean (0.67) would have about 12.6% of modern sector workers employed in the civil service. A one standard deviation increase in ethnic diversity (from 0.67 to 0.95) correlates with a 3.5% increase in the size of the civil service. This is close to half a standard deviation in the dependent variable. The magnitude of the effect is consistent across years. Part of the correlation in Models 1-4 is a product of the close relationship between ethnic diversity and population size. Diverse provinces tend to have smaller populations, and provinces with smaller populations tend to have more civil servants. Models 5-8 add in a provincial population variable (logged). Inclusion of the population variable changes the performance of the model considerably. On average, Models 1-4 explain 30% of variation in the dependent variable. In contrast, Models 5-8 explain 51% of variation in civil service size. Additionally, the inclusion of the population variable produces a substantial decrease in the size of the ethnic fractionalization coefficients. On average, the ethnic fractionalization coefficients are 36% smaller when the population controls are included. This is a substantively large difference. In the year 2005, for instance, the model now predicts that a one standard deviation in ethnic fractionalization correlates with a 2.4% increase in civil service size. This indicates that a substantial portion of the correlation between ethnic fractionalization and civil service size was simply a product of population size. Nonetheless, a positive and statistically significant relationship remains even when population controls are added. Population size only explains a part of the relationship. Section 2 – Transfers and Civil Service Size The correlation between civil service size and ethnic diversity is partially attributable to the fact that diverse sub-national units receive higher per-capita transfers from the centre. Large civil services are funded with the large transfers. The correlation between transfers and diversity, however, is purely a product of population size. In this section I establish the plausibility of both claims. Dependent Variable: Percentage of modern sector workers employed in the civil service  From the 2005 Statistical Yearbook. See Section 1. Independent Variable: Pre capita transfers (provincial)  Total per capita transfers to sub-national governments (2005): This variable was constructed by adding the total flow of DAU and DAK to a province. The total transfers were divided by the population of the province. This value provided the per capita transfers in thousands of Rupiah. This value was then logged. 259 Figure 33 - Appendix C – Civil service size and transfers There is a clear, positive correlation between civil service size and transfer flows. Those provinces that take in the greatest per-capita transfer payments have the largest civil services. Note that the outlier, with very low transfers and a modest civil service size, is the capital city of Jakarta, where many national-level bureaucrats live. Section 3 – Transfers and Ethnic Diversity Dependent Variable: Per capita transfers (municipal)  Transfers per municipal unit: This variable combined the total DAK and DAU transfers to the municipal unit in 2005. It was constructed using a method similar to that in Section 2. The transfers were aggregated so that they matched the municipal boundaries of 2000 as to better match the ethnicity data. In some cases this meant combining two or more flows of transfers. The use of the 2005 data was justified by the fact that the available Indonesian data tends to be most detailed after 2003. As well, using 2005 allows re-use of the same data collected during the test in Section 2. Independent Variables  Municipal Ethnic Fractionalization - 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous. Measured at the municipal level.  Population Size - Total population living within the municipality (logged) 255 255 Population data was also drawn from the 2000 census. 260 Results Model 1 tests only the relationship between ethnic diversity and transfers. A statistically significant relationship exists. Diverse municipalities tended to receive higher volumes of per-capita transfers than homogenous municipalities. Of course, the measurement is complicated by the process of data generation, but the relationship is strong and statistically significant. Model 2 adds in the population data for each municipality. Municipalities with large populations tended to receive less per capita transfers than municipalities with small populations. This is intuitive and consistent with the practice of municipal splitting. Indeed, logged population size explains most of the variation in the dependent variable. The relationship between ethnic diversity and transfers slips below statistical significance. Diverse municipalities do not receive more transfers because they are diverse; rather, diverse municipalities receive more transfers because they are small. In sum: At the provincial level, the correlation between transfers and civil service size strongly suggests that large civil services tend to be funded through high per capita transfer payments. A close examination of municipal level transfers reveals that diverse municipalities do tend receive more per capita funds, but that the relationship is purely an artifact of population size. Table 28 - Appendix C – Transfers, diversity, and population size Per-Capita Transfers to Municipalities – 2005 DV: Thousands of Rupiah per person, logged Model 1 - Ethnicity Model 2 – Population Variable Estimate (std. err) Estimate (std. err) Ethnic Fractionalization (Municipality) .65*** (.11) .05 (.05) Population (logged) -.67*** (.02) const 6.11*** (.06) 14.92*** (.23) Observations 333 333 R2 0.0916 0.8325 *p < .10. **p < .05. *** p <.01. 261 Section 4 - Corruption perceptions and ethnic diversity Figure 6 presents a suggestive correlation between municipal corruption perceptions and ethnic diversity. This section demonstrates the robustness of the relationship. Using Transparency International Indonesia data from 2008 and 2010, I establish a strong correlation between corruption and ethnic diversity even when other important demographic and economic variables are controlled. Dependent Variable: Corruption Perception  The TI Indonesian Corruption Index provides one measure of corruption perceptions on Indonesia. It focused on 50 municipalities across the country. One disadvantage is that it only covers cities. This being said, it still provides a measure to work with. Results are presented in a 1-10 scale, 1 being the most corrupt and 10 being the least. I use data from two years: 2008 and 2010. Additionally, I combine the two values to find an average corruption perception across time. I use all three values – 2008, 2010, and the average – as dependent variables. Independent Variables  Municipal Ethnic Fractionalization - 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous. Measured at the municipal level.  Population Size - Total population living within the municipality (logged) 256  Per-Capita Income - Per-Capita municipal GDP logged Results Ethnic fractionalization strongly correlates with corruption perceptions in both 2008 and 2010. High fractionalization scores correlates with low TI scores. In other words, diverse municipalities were consistently rated the most corrupt. The sizes of the correlation coefficients are consistent across time. Notably the most variation in the dependent variable is explained in Model 3, which averages the two TI scores. Population size also correlates with corruption perceptions. Municipalities with small population sizes were consistently ranked as less corrupt than large cities. Surprisingly, there is no correlation between income levels and corruption perceptions. Simulations help demonstrate the size of the effect. With all three variables set to the mean, the predicted corruption perception score is 4.47 in 2008 and 4.93 in 2010. An increase of one standard deviation in ethnic fractionalization produces a predicted change 256 From the “Province in Numbers” Series. For reasons of data availability, the same independent variable values are used for all tests. Though this does miss changes across time, these minor differences should not substantively effect the results. 262 of 0.26 and 0.30 in 2008 and 2010 respectively. In both years, the magnitude of the effect is over one-third of a standard deviation in the dependent variable. A test of the average scores (Model 3) yields similar results. In this case, a change in one standard deviation in ethnic fractionalization correlates with a corruption score decline of over one-half of a standard deviation in the dependent variable. In sum: ethnic fractionalization strongly correlates with municipal level corruption perceptions. The finding is consistent across time. The magnitude of the effect is substantively large, with a one standard deviation increase in the fractionalization variable producing between a decline in the corruption score of somewhere between one- third and one-half of a standard deviation. Table 29 - Appendix C – Determinants of corruption Transparency International Corruption Perception Index 2008, 2010 (1-10) Model 1 - TI 2008 Model 2 – TI 2010 Model 3 - Average TI Variable Estimate (std. err) Estimate (std. err) Estimate (std. err) Ethnic Fractionalization (Municipality) -.86** (.35) -.99*** (.31) -.92*** (.22) Population (logged) -.20* (.11) -.21** (.10) -.20*** (.07) Income (logged) .11 (.13) .001 (.11) .05 (.08) const 6.56*** (1.72) 8.27*** (1.50) 7.42*** (1.08) Observations 50 50 50 R2 0.0791 0.2098 0.3066 *p < .10. **p < .05. *** p <.01. Section 5 - Infrastructure quality and ethnic diversity Figure 7 displays a strong correlation between public service provision and ethnic diversity. This Appendix tests the robustness of the relationship through multivariate regressions. It breaks down the unit of analysis to the municipal level. As well, it tests the consistency of the relationship over time through an analysis of service provision in 2003. Dependent Variable: KPPOD Infrastructure Scores 1. KPPOD Local Infrastructure Sub-Index 2007 - KPPOD’s 2007 municipal infrastructure rating is based on evaluations made by private sector actors. Respondents rank the municipality’s ability to deliver 5 core infrastructure services (electricity, roads, street illumination, water, and phone). Additionally, 263 respondents provide information on average road repair times, public electricity provision, and firm-based generator use. KPPOD uses all of this data to create a Local Infrastructure Sub-Index. The KPPOD sample included 12,187 firms from 243 municipalities in 15 provinces. While not all municipalities are covered, the sample provides a range of provinces, with at least one province covered on each of the major islands (Sumatra, Java, Kalimantan, and Sulawesi). In 2007, the infrastructure score was reported on a scale of 0-100. 2. Dependent Variable 2: KPPOD Quality of Roads Scores in 2003: The 2003 dataset was the second year KPPOD offered comprehensive municipal measurements of performance. Using the 2003 dataset provides a snapshot of municipal performance at the beginning of both the democratic era and decentralization. Because the breadth of municipalities covered significantly improved between 2002 and 2003 (from 132 to 200), the 2003 data is more amendable to multivariate statistical analysis. In 2003, KPPOD offered a range of infrastructure measures generated through both surveys and expert analysis. KPPOD’s “Quality of Roads” (“Kualitas jalan”) rating offers one potentially testable proxy for infrastructure delivery that falls under the jurisdiction of the municipality. Maintaining road quality is a constant challenge that requires continual work by authorities. Politicians concerned with maintaining clientelistic linkages may choose to squander road budgets in various ways, including the diversion of funds into private coffers or passing out repair contracts based on political connections rather than ability to deliver the service. On the other hand, the ability to effectively maintain roads reflects a concern for public goods delivery on the part of politicians. KPPOD transformed their road data to reflect both the intensity of responses and the contribution a particular variable makes to a larger investment climate index. This process makes the data analysis difficult to interpret in concrete terms. Not accounting for the weighting process done by KPPOD, the publically available data suggests there are four relative performance categories that can translate roughly into High, Moderate, Low and Extremely Low. For simplification, I transform the KPPOD measure into a 1-4 performance measure, 1 being the worst performance and 4 being the best. Independent Variables  Municipal Ethnic Fractionalization - 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous. Measured at the municipal level  Provincial Ethnic Fractionalization - 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous. Measured at the provinicial level 264  Population Size - Total population living within the municipality (logged) 257  Per-Capita Income - Per-Capita municipal GDP logged258  Population Density – Total municipal population / total geographic landmass (Km2). Logged.  Effective Number of Electoral Parties – Effective Number of Electoral Parties in the Municipality. Results Table 30 presents Results from 2007 . Models 1 and 2 include bivariate tests using two different ethnic fractionalization measures. In both Models the ethnic fractionalization variables strongly and negatively correlate with the Local Infrastructure Sub-Index. Notably, the provincial fractionalization measure (Model 2) explains slightly more variation than the municipal fractionalization measure. Model 3 adds the social-economic control variables. Only population density correlates with infrastructure provision. As one would expect, densely populated municipalities have higher service provision scores. Both ethnic fractionalization variables remain negative and statistically significant. The provincial ethnic fractionalization measure has the strongest correlation with service provision, with a correlation coefficient almost three times the size of the municipal level fractionalization variable. Additionally, the relationship is significant at the P<.01 level. These results remain difficult to interpret, however, due to significant colinearity between the two key ethnic variables. Model 4 adds the effective number of electoral parties. The inclusion of the variable causes the municipal-level ethnic fractionalization to lose statistical significance. The effective number of electoral parties, however, appears strongly significant, which suggests these two variables are related. Still, the provincial fractionalization measure remains strongly significant. To demonstrate the size of the effect I ran simulations of Model 4. Starting from the mean provincial ethnic fractionalization (0.53), a one standard deviation increase (0.31) in the value of the variable caused the predicted infrastructure score to decrease from 65.6 to 61.7. The magnitude of the change in the infrastructure variable is equivalent to a decrease of one-third (31%) of a standard deviation in the dependent variable. Table 31 presents results from 2003. Model 1 and 2 present stripped down tests of only municipal and provincial fractionalization measures. While both are statistically significant in stand alone models, the provincial data provides far more leverage when 257 Each provincial branch of the Central Bureau of Statistics releases an annual publication, typically titled “Provinisi X Dalam Angka” or “Province X in Numbers.” While complicating replication, I used these sources because Jakarta based publications of the aggregate provincial data typically lag behind annual provincial data. 258 Specifically, I took the data from Produk Domestik Regional Bruto Kabupaten/Kota Di Indonesia 2003- 2007. While the volume does contain income for both 2006 and 2007, the figures are flagged as preliminary. Consequently, I used income figures from 2005. 265 accounting for variation across municipalities. Model 3 includes both measures and all three socio-economic controls. Municipal fractionalization is no longer statistically significant, while provincial fractionalization continues to have the predicted negative effect. Surprisingly, the only control variable that reaches significance is population density. Dense municipalities tend to be given much higher infrastructure scores. Model 4 adds in the effective number of electoral parties. The variable has a slight, negative impact on road quality, though the correlation does not reach statistical significance. Again, I ran simulations to demonstrate the size of the effect. With all values set to the mean, the predicted Road Quality score was 3.01. A one standard deviation increase in the provincial fractionalization variable, from 0.57 to 0.87, resulted in a predicted decrease in the Road Quality variable of 0.29. The magnitude of the predicted erosion in Road Quality is approximately one-third of a standard deviation in the dependent variable. This is only a rough test, however, as such a change is impossible in an ordinal variable such as Road Quality. To conclude, tests of KPPOD’s infrastructure data show that provision of infrastructure services is consistently worse in ethnically diverse provinces. This affect is independent of a range of other socio-economic and political variables. The only socio-economic factor that was consistently significant across all relevant tests in all years was population density. A system with a high number of effective parties is correlated with poor infrastructure delivery in several models. It is, however, difficult to tell if this is a wholly independent affect. The 2003 KPPOD data was collected in 2002. This only allows two full years – 2000 and 2001 – in which the party system could have produced the infrastructure outcomes analyzed in the models. In all likelihood party systems are being shaped by the same clientelistic distributive patterns that drive the infrastructure results. 266 Table 30 - Appendix C – Determinants of infrastructure provision [2007] KPPOD 2007 Infrastructure Sub-Index (1-100) Model 1 - Simple District Fractionalization Model 2 - Provincial Fractionalization Model 3 - Socio-Economic Controls Model 4 - Political Controls Variable Estimate (std. err) Estimate (std. err) Estimate (std. err) Estimate (std. err) Ethnic Fractionalization (District) -20.05*** (2.033) -5.74* (3.13) -3.71 (3.16) Ethnic Fractionalization (Province) -23.35*** (2.02) -14.15*** (3.23) -12.51*** (3.23) Population (logged) -.65 (.75) -.79 (.74) Income (logged) -1.01 (.87) -.91 (.85) Population Density (logged) 1.66*** (.43) 1.76*** (.43) Electoral Parties -.82*** (.28) const 73.04*** (1.00) 78.03*** (1.25) 82.81*** (12.50) 87.21*** (12.39) Observations 243 243 243 243 R2 0.28 0.35 0.40 0.43 *p < .10. **p < .05. *** p <.01. 267 Table 31 - Appendix C - Determinants of infrastructure provision [2003] KPPOD 2003 District Road Quality (1-4) Model 1 - Simple District Fractionalization Model 2 - Provincial Fractionalization Model 3 - Socio-Economic Controls Model 4 - Political Controls Variable Estimate (std. err) Estimate (std. err) Estimate (std. err) Estimate (std. err) Ethnic Fractionalization (District) -1.23*** (.19) -.32 (.28) -.30 (.29) Ethnic Fractionalization (Province) -1.68*** (.20) -.95*** (.31) -.98*** (.31) Population (logged) -.06 (.07) -.04 (.07) Income (logged) .13 (.09) .14 (.09) Population Density (logged) .15*** (.04) .15*** (.04) Electoral Parties -.05 (.04) const 3.61*** (.10) 4.05*** (.13) 1.69 (1.67) 1.50 (1.67) Observations 199 200 198 197 R2 0.1716 0.2660 0.3529 0.3608 *p < .10. **p < .05. *** p <.01. 268 Appendix D – Supplement to Chapter 4 Section 1 I use two criteria for categorization of Muslim parties: 1) official pronouncements denoting ‘Islam’ as the basis of the party; 2) partisan origins traceable to pre-existing religious organizations. The first criteria captures explicitly ‘Muslim’ parties, the second includes ostensibly secular parties which have origins in the Muslim social organizations. In denoting parties ‘Muslim’ I erred on the side of inclusion. Below I include a full list of off ‘Muslim’ parties for all three elections. In addition to this dummy, a variable capturing the interaction between the percentage of Muslims and Muslim Party is included. Muslim Parties For 1999: 1) Partai Kebangkitan Umat; 2) Partai Nahdlatul Ummat; 3) Partai Kebangkitan Bangsa; 4) Partai Solidaritas Uni Nasional Indonesia; 5) Partai Indonesia Baru; 6) Partai Amanat Nasional; 7) Partai Islam Demokrat; 8) Partai Kebangkitan Muslim Indonesia; 9) Partai Ummat Islam; 10) Partai Masyumi Baru; 11) Partai Persatuan Pembangunan; 12) Partai Syarikat Islam Indonesia; 13) Partai Abul Yatama; 14) Partai Syarikat Islam Indonesia 1905; 15) Partai Politik Islam Indonesia Masyumi; 16) Partai Bulan Bintang; 17) Partai Keadilan; 18) Partai Persatuan; 19) Partai Daulat Rakyat; 20) Partai Cinta Damai; 21) Partai Ummat Muslimin Indonesia. For 2004: 1) Partai Bulan Bintang; 2) Partai Persatuan Pembangunan; 3) Partai Persatuan Nahdlatul Ummah Indonesia; 4) Partai Amanat Nasional; 5) Partai Kebangkitan Bangsa; 6) Partai Keadilan Sejahtera; 7) Partai Bintang Reformasi. For 2009: 1) Partai Keadilan Sejahtera; 2) Partai Amanat Nasional; 3) Partai Kebangkitan Bangsa; 4) Partai Matahari Bangsa; 5) Partai Persatuan Pembangunan; 6) Partai Bulan Bintang; 7) Partai Bintang Reformasi; 8) Partai Kebangkitan Nasional Ulama; 9) Partai Persatuan Nahdlatul Ummah Indonesia. 269 Section 2 In Indonesia, parties that fail to meet a threshold of either aggregate national or sub- national strength can not automatically re-offer the following election. If a party fails to meet the threshold it must either present itself under a new name and complete the full registration process again, or amalgamate with other parties so that they can meet the threshold, thereby avoiding the more extensive registration process. Each post-1999 election, then, produces four possible categories of parties: 1) carry-overs from the previous election; 2) re-named parties from the previous election; 3) an amalgamation of parties from the previous election; 4) a completely new party. For the construction of the electoral strength variables, both re-named and amalgamated parties were assigned the electoral vote of their previous incarnation(s). Listed below is a full accounting of all parties that ran in two consecutive elections. Party History 2004: Carry-Overs: 1) Partai Bulan Bintang; 2) Partai Persatuan Pembangunan; 3) Partai Amanat Nasional; 4) Partai Kebangkitan Bangsa; 5) Partai Demokrasi Indonesia Perjuangan; 6) Partai Golongan Karya (Golkar). Name Changes: 1) Partai Nasional Indonesia Marhaenisme (formerly Partai Nasional Indonesia); 2) Partai Buruh Sosial Demokrat (formerly Partai Buruh Nasional); 3) Partai Keadilan dan Persatuan Indonesia (formerly Partai Keadilan Dan Persatuan);4) Partai Penegak Demokrasi Indonesia (formely Partai Demokrasi Indonesia); 5) Partai Persatuan Nahdlatul Ummah Indonesia (formely Partai Nahdlatul Ummat); 6) Partai Keadilan Sejahtera (formerly Partai Keadilan). Amalgamations: 1) Partai Bintang Reformasi (Partai Indonesia Baru + Partai Ummat Muslimin Indonesia + Partai Kebangkitan Muslim Indonesia + Partai Republik); 2) Partai Sarikat Indonesia (Partai Syarikat Islam Indonesia + Partai Daulat Rakyat + Partai Politik Islam Indonesia Masyumi). 2009: Carry-Overs: 1) Partai Karya Peduli Bangsa; 2) Partai Keadilan dan Persatuan Indonesia; 3) Partai Keadilan Sejahtera; 4) Partai Amanat Nasional; 5) Partai Persatuan Daerah; 6) Partai Kebangkitan Bangsa; 7) Partai Nasional Indonesia Marhaenisme; 8) Partai Penegak Demokrasi Indonesia; 9) Partai Pelopor; 10) Partai Golongan Karya; 11) Partai Persatuan Pembangunan; 12) Partai Damai Sejahtera; 13) Partai Bulan Bintang; 14) Partai 270 Demokrasi Indonesia Perjuangan; 15) Partai Bintang Reformasi; 16) Partai Demokrat; 17) Partai Merdeka; 18) Partai Sarikat Indonesia; Name Changes: 1) Partai Perjuangan Indonesia Baru (formerly Partai Perhimpunan Indonesia Baru); 2) Partai Demokrasi Kebangsaan (formerly Partai Persatuan Demokrasi Kebangsaan); 3) Partai Nasional Benteng Kerakyatan Indonesia (formerly Partai Nasional Banteng Kemerdekaan); 4) Partai Patriot (formerly Partai Patriot Pancasila); 5) Partai Nahdlatul Ummah Indonesia (formerly Partai Persatuan Nahdlatul Ummah Indonesia); 6) Partai Buruh (formerly Partai Buruh Sosial Demokrat Section 3 Table 32 - Appendix D – Determinants of entry [1999-2009] Determinants of Candidate List Size – National Legislature (OLS Regression, Clustered by Party) Dependent Variable: Party-level Candidates-per-Seat by Electoral District Model 1 – 1999 Model 2 - 2004 Model 3 – 2009 Model 4 - 2004 Variable Estimate (std. err) Estimate (std. err) Estimate (std. err) Estimate (std. err) Ethnic Fractionalization .087 (.069) .112*** (.024) .111*** (.021) .113*** (.024) District Magnitude -.003*** (.0009) -.036*** (.003) -.024*** (.003) .036*** (.003) Jakarta .206* (.114) .145*** (.034) .126*** (.020) .144*** (.034) Urbanization .070 (.189) .126*** (.031) .122*** (.022) .125*** (.031) Poverty -.035 (.233) -.133** (.060) .0484 (.068) -.133** (.058) Muslim Party -.197*** (.060) -.193 (.126) -.200* (.099) -.196* (.105) % Muslim .122 (.074) -.089** (.037) -.175*** (.034)) -.080** (.036) % Muslim X Muslim Party .304*** (.068) .434*** (.072) .463*** (.091) .401*** (.078) National Strength 1.771*** (.554) Local Strength .441*** (.149) const .553*** (.121) .848*** (.070) .711*** (.066) .762*** (.055) Obs 1248 1656 2926 1656 R2 0.0326 0.1775 0.1149 0.4814 *p < .10. **p < .05. *** p <.01. 271 Section 4 Table 33 - Appendix D – Determinants of provincial entry Determinants of Candidate Numbers – Provincial Legislature (OLS Regression) Dependent Variable: Candidates-per-Seat (Total Candidates / Total Seats) Variable Model 1 – 2004 Model 2 - 2009 Model 3 - 2004 Model 4 - 2009 Estimate (std. err) Estimate (std. err) Estimate (std. err) Estimate (std. err) Ethnic Fractionalization 3.843** (1.526) 6.742*** (1.863) 4.108*** (1.369) 6.385*** (1.828) DPRD Seats .022 (.020) .035 (.027) Jakarta 8.873*** (2.690) 10.580*** (3.667) Urbanization -4.236 (3.157) -6.336 (4.130) Poverty -5.809* (3.270) 3.094 (6.121) Aceh -2.170 (2.737) Const 10.197*** (1.106) 11.896*** (1.358) 11.498*** (2.329) 11.546*** (3.141) Obs 32 33 32 33 R2 0.1745 0.2971 0.5384 0.5260 *p < .10. **p < .05. *** p <.01. Section 5 Table 34 - Appendix D – Determinants of upper house entry Determinants of Candidate Numbers – National Upper House (OLS Regression) Dependent Variable: Number of Candidates Model 1- 2004 Model 2 – 2009 Model 3 – 2004 Model 4 – 2009 Variable Estimate (std. err) Estimate (std. err) Estimate (std. err) Estimate (std. err) Ethnic Fractionalization -17.143** (6.902) 5.278 (8.457) -3.413 (4.738) 8.274 (7.943) Population (logged) 8.123*** (1.266) -.736 (3.152) Urbanization -2.280 (8.514) -13.452 (14.290) Poverty -19.553* (11.456) -43.901 (27.202) Electoral Fragmentation, 2004 1.337** (.565) const 40.762*** (5.001) 29.830*** (6.164) -85.368*** (21.071) 30.846 (43.573) N 32 33 32 32 R2 0.1706 0.0124 0.7086 0.4152 *p < .10. **p < .05. *** p <.01. 272 Section 6 Table 35 - Appendix D – Determinants of entry by party size Determinants of Candidate List Size – National Legislature (OLS Regression, Clustered by Party) Dependent Variable: Party-level Candidates-per-Seat by Electoral District Variable Model 1 – Minor Parties 2004 Model 2– Minor Parties 2009 Model 3 – Major Parties 2004 Model 4 – Major Parties 2009 Model 5– Homogenous Districts, Minor Parties Model 6 – Homogenous Districts, Major Parties Variable Estimate (std. err) Estimate (std. err) Estimate (std. err) Estimate (std. err) Estimate (std. err) Estimate (std. err) Rent Opportunities .005** (.002) .003 (.003) Ethnic Fractionalization .122*** (.028) .091*** (.022) .044 (.053) .106 (.055) District Magnitude -.038*** (.003) -.025*** (.003) -.029** (.008) -.018** (.006) -.025*** (.005) -.002 (.013) Jakarta .151*** (.041) .123 (.060) Urbanization .159*** (.032) .145*** (.025) .026 (.070) .075* (.034) .043 (.060) .096 (.103) Poverty -.075 (.063) .016 (.071) -.287* (.132) .186 (.131) -.300 (.177) .126 (.422) Muslim Party -.332** (.123) -.180 (.152) -.560*** (.084) -.524*** (.059) % Muslim -.081* (.040) -.178*** (.034) .0002 (.080) .090 (.049) Muslim Party X % Muslim .474*** (.094) .348** (.139) .148 (.109) .222 (.128) National Strength .869 (5.39) -8.6457* (4.566) -1.677*** (.219) -2.12** (.851) -4.874 (4.950) .467 (.503) Local Strength 3.63*** (1.21) 3.698*** (.819) .341 (.175) .923** (.274) -1.427 (2.131) .341 (.328) .737*** (.060) .672*** (.072) 1.762*** (.137) 1.299*** (.091) Const .499*** (.068) .789** (.219) Obs 1311 2294 345 518 589 133 R2 0.2453 0.1236 0.3942 0.4127 0.0532 0.0559 *p < .10. **p < .05. *** p <.01. 273 Section 7 Table 36 - Appendix D – Entry across levels of governance Determinants of Candidate List Size – National Legislature (OLS Regression, Clustered by Party) Dependent Variable: Party-level Candidates-per-Seat by Electoral District Variable Model 1 – Full Sample Model 2 – Minor Parties Model 3 – Major Parties Estimate (std. err) Estimate (std. err) Estimate (std. err) DPRD Candidates-per- seat .514*** (.079) .538*** (.083) .056 (.076) Ethnic Fractionalization .057** (.020) .060** (.025) .041 (.049) District Magnitude -.040*** (.003) -.043*** (.003) -.029*** (.008) Jakarta -.015 (.046) -.026 (.050) .109 (.071) Urbanization .150*** (.026) .174*** (.029) .031 (.067) Poverty -.100* (.055) -.061 (.060) -.278* (.130) % Muslim -.028 (.035) -.017 (.037) -.005 (.079) Muslim Party -.087 (.088) -.197* (.092) -.537*** (.078) % MuslimXMuslim Party .184** (.071) .224* (.097) .139 (.103) National Strength .975** (.374) 1.059 (3.362) -1.640*** (.259) Local Strength .096 (.146) 1.828 (1.280) .307 (.173) Const .556*** (.064) .536*** (.061) 1.704*** (.207) Obs 1656 1311 345 R2 0.5816 0.4093 0.3959 *p < .10. **p < .05. *** p <.01. 274 Appendix E – Supplement to Chapter 5 Section 1 Coding criteria for matched names Same Name?259 If Yes, Proceed Step 2a If No, Proceed Step 2b 2a. Complex Name?260 2b. Near Match?261   If Yes, Code as Match If No, Proceed to Step 3 If Yes, Proceed If No, discard  3. Same Party?262  If Yes, Code as Match If No, Proceed  4. Incongruent Title?263  If Yes, Discard If No, Proceed  259 Criteria: 100% match 260 Criteria: 1) Over 8 letters (e.g. Zulkarnaen); 2) Not a common Muslim name (e.g. Muhammad Anwar) 261 Criteria: 1) similar phonetics (e.g. Muhammad Anwar vs. Muhamad Anwar); 2) small missing element (Muhammad Anwar Ali vs. Muhammad Anwar) 262 Criteria: 1) exact same party: 2) renamed party; 3) amalgamated party 263 Criteria: 1) Obvious disjuncture between significant elements (e.g. Kol (purn) Anwar Ali vs. Drs. Anwar Ali) 275 5. Matching Title?264  If Yes, Code as Match If No, Proceed  6. Matching Home Area?265  If Yes, Code as Match, If No, Discard 264 Criteria: 1) Exact match; 2) At least one major element (e.g. Drs. H. Anwar Ali vs. Drs. Anwar) 265 Criteria: 1) Same province 276 Section 2 I add the TI score to the second mode in Table 10. As an interpretive convenience, I subtract the score from 10 so perceived corruption is worse in those provinces with higher scores. Additionally, I add an interactive variable, as the effect of low constraints (corruption) should be conditional on access to resources (civil service size). Results are displayed below. Standard errors on interactive terms are difficult to interpret. I follow Brambor, Clark & Golder’s (2006) method to generate a figure demonstrating the conditional effect of corruption. The figure captures the marginal effect of a one-unit change in the TI variable, dependent upon civil service size. The Y- Axis measures the expected change in the relative proportion of Switchers to Loyalists. What we find is a slope that rises with corresponding increases in civil service size. In other words, corruption induces switching in those provinces with large bureaucracies. This is consistent with the rent opportunities story: it is the combination of low constraints and high resource access that modifies behaviour. Despite the fact the slope is consistent with my hypothesis, the interactive term falls slightly below standard levels of statistical significance. From a civil service size of 0.125 (12.5%), the lower 90% confidence interval skirts a value of 0. This indicates that we can not confidently predict that a 1-unit change in the TI value will produce increasing rates of party switching. The uncertainty that we see at higher levels of civil service size likely reflects a small number observations. The interactive term is suggestive and adds a small measure of support for the rent opportunities story. Table 37 - Appendix E – An interactive model of party switching Proportion of Switchers Across Districts Dependent Variable: Switchers / Run-Agains Time Period: 2004- 2009 Variable Estimate (std. err.) Civil Service Size -4.98 (5.22) Provincial Magnitude 0.002* (0.001) Corruption (Avg TI score) -0.08 (0.11) CivilServiceXCorruption 1.14 (0.98) Const 0.46 (0.57) N 32 R2 0.2493 *p < .10. **p < .05. *** p <.01. 277 Figure 34 - Appendix E – Marginal effect of corruption on party switching 278 Appendix F – Supplement to Chapter 6 Section 1 In most sub-national units, transfers from the central government constitute the vast majority of revenues. Transfers are not, however, equal across the country. On a per- capita basis, some areas receive more transfers than others. I thus use per-capita transfers as a variable to capture the resources available to politicians. To construct the variable I use transfer data from 2004. Two streams of transfers were included: DAU and DAK. I aggregate the total DAK and DAU funds transferred to the municipal units within each electoral district. To this total I add an electoral districts share of transfers made to the provincial unit. I then use the 2000 Census data to create a per-capita measure of transfers. The transfer variable yields qualitatively similar results. Starting from the mean, a one standard deviation increase in transfer funds results in an approximately 4.6% increase in preference voting. Table 38 – Appendix F – Transfers and preference voting Rent Opportunities and Preference Voting (OLS) Model 1 Estimate (std. err) Per Capita Transfers(logged) 7.01*** (0.45) District Magnitude 0.78*** (0.17) Urbanization 0.13*** (0.02) Poverty 0.09*** (0.02) Ethnic Fractionalization 12.95*** (1.11) Candidate List Size 0.22*** (0.02) Home Candidates 0.05** (0.02) Women Candidates 0.008 (0.02) Constant -30.86*** (4.80) Observations 1646 R2 0.4825 *p < .10 **p < .05 *** p <.01 279 Section 2 Table 12 presents results without any district level fixed effects. Below I presents results with series of district level dummies are added to the model. Though a few of the control variables are affected by the alternative specification, the core independent variables remain positively signed and strongly significant. Indeed, if anything the simple model presented in the body text underestimates the effect of rent opportunities on preference voting. Table 39 - Appendix F –Fixed effects model of preference voting Rent Opportunities and Preference Voting (OLS, District Dummies) Model 1 Model 2 Model 3 Estimate (std. err) Estimate (std. err) Estimate (std. err) Civil Service Size (2005) 1.95*** (0.15) Civil Service Size (1999) 1.93*** (0.16) Transfers (by pop) 4.81*** (1.25) District Magnitude -0.28 (0.26) 0.13 (0.44) Urbanization 0.009 (0.03) 0.06* (0.03) 0.28*** (0.05) Poverty -0.30*** (0.06) -0.36*** (0.06) 0.58*** (0.08) Ethnic Fractionalization 4.66 (2.98) -3.29 (3.37) 15.25*** (2.80) Candidate List Size 0.20*** (0.02) 0.20*** (0.02) 0.20*** (0.02) Home Candidates 0.06** (0.02) 0.06** (0.02) 0.06** (0.02) Women Candidates 0.002 (0.02) 0.002 (0.02) 0.002 (0.02) Constant 22.34*** (3.72) 15.52*** (4.83) -28.76*** (10.21) Observations 1646 1646 1646 R2 0.6532 0.6532 0.6532 *p < .10 **p < .05 *** p <.01 280 Section 3 Table 12 presents results with standard errors clustered by party. To further test the robustness of the key relationship across parties I re-ran the model for each individual party. Despite the severe restriction in the number of observations (maximum: 69), coefficients for the Rent Opportunity variables were all positively signed and, with a small number of exceptions, statistically significant. The table below presents the results by party. In the interests of space I only present coefficients and standard errors on the key Rent Opportunity variables. Table 40 - Appendix F – Effect of rents by party Relationship Between Rent Opportunities and Preference Voting by Party Party Name Pct Civil Service (2005) Pct Civil Service (1999) Transfers (logged) Estimate (std. err) Estimate (std. err) Estimate (std. err) Partai Nasional Indonesia Marhaenisme (PNI-M) 0.83*** (0.29) 0.98*** (0.28) 5.84** (2.55) Partai Buruh Sosial Demokrat (PBSD) 0.70** (0.30) 0.94*** (0.30) 7.18** (2.73) Partai Bulan Bintang (PBB) 0.72** (0.29) 0.98*** (0.28) 5.13* (2.73) Partai Merdeka (PM) 1.39*** (0.34) 1.07*** (0.37) 8.63** (3.22) Partai Persatuan Pembangunan (PPP) 0.79*** (0.24) 0.79*** (0.24) 4.12* (2.32) Partai Persatuan Demokrasi Kebangsaan (PDK) 1.28*** (0.31) 1.46*** (0.31) 5.31 (3.22) Partai Perhimpunan Indonesia Baru (PIB) 1.22*** (0.37) 1.58*** (0.37) 7.71** (3.53) Partai Nasional Banteng Kemerdekaan (PNBK) 1.14*** (0.32) 1.10*** (0.35) 8.69*** (2.99) Partai Demokrat (PD) 0.89*** (0.24) 0.79*** (0.24) 6.37*** (2.24) Partai Keadilan dan Persatuan Indonesia (PKPI) 1.33*** (0.31) 1.23*** (0.31) 10.95*** (2.89) Partai Penegak Demokrasi Indonesia (PPDI) 1.74*** (0.33) 1.28*** (0.36) 11.96*** (3.16) Partai Persatuan Nahdlatul Ummah Indonesia (PPNUI) 1.11*** (0.35) 1.13*** (0.39) 6.11* (3.16) Partai Amanat Nasional (PAN) 0.76*** (0.23) 0.84*** (0.22) 6.59*** (2.17) Partai Karya Peduli Bangsa (PKPB) 1.18*** (0.32) 1.12*** (0.31) 6.64** (2.71) Partai Kebangkitan Bangsa (PKB) 0.82**** (0.26) 0.91*** (0.26) 3.23 (2.45) Partai Keadilan Sejahtera (PKS) 0.74*** (0.18) 0.64*** (0.19) 4.57** (1.76) Partai Bintang Reformasi (PBR) 1.10*** (0.28) 1.15*** (0.29) 6.86** (3.17) 281 Party Name Pct Civil Service (2005) Pct Civil Service (1999) Transfers (logged) Partai Demokrasi Indonesia Perjuangan (PDI-P) 1.50*** (0.31) 1.05*** (0.35) 7.09** (3.34) Partai Damai Sejahtera (PDS) 0.70** (0.29) 0.93*** (0.29) 5.28* (2.64) Partai Golongan Kaya (Golkar) 1.39*** (0.28) 1.25*** (0.28) 7.80*** (2.81) Partai Patriot Pancasila (PP) 1.00*** (0.32) 1.25*** (0.32) 6.50** (3.10) Partai Sarikat Indonesia (PSI) 1.44*** (0.30) 1.31*** (0.29) 9.80*** (2.69) Partai Persatuan Daerah (PPD) 1.03*** (0.39) 1.24*** (0.40) 8.92** (3.42) Partai Pelopar (PP) 1.05*** (0.36) 1.46*** (0.39) 6.80** (3.20) *p < .10 **p < .05 *** p <.01 Section 4 It is plausible that small population size could be independently causing both high levels of rent opportunities and preference voting. To investigate the possibility I added two distinct population variables. The first captures the total population within the electoral district. The second captures the total population within the province. The logic of the second test is that small provinces necessarily face large start up costs in order to staff provincial branches of various departments. Thus a low population district in a low population province should naturally have more state employees than a low population district in a high population province. Adding the provincial level population variable allows me to parse out the effect of rent opportunities from the potential quirks that accompany small province politics. All population data was drawn from the 2004 census. Table 41 presents results controlling for district population size. Table 42 presents results controlling for provincial population size. Inclusion of the variables does not alter the substantive findings of Table 12. 282 Table 41 - Appendix F – Preference voting and district population size Rent Opportunities and Preference Voting – Population Test #1 (OLS) Model 1 Model 2 Model 3 Estimate (std. err) Estimate (std. err) Estimate (std. err) Civil Service Size (2005) 1.07*** (0.08) Civil Service Size (1999) 1.23*** (0.09) Transfers (by pop) 6.53*** (0.84) District Population (logged) 0.71 (0.83) 3.21*** (1.12) -1.03 (1.34) District Magnitude 0.64*** (0.20) 0.13 (0.23) 0.90*** (0.20) Urbanization 0.09*** (0.02) 0.07** (0.02) 0.12*** (0.02) Poverty -0.05* (0.02) -0.03 (0.02) 0.09*** (0.02) Ethnic Fractionalization 5.42*** (0.95) 6.45*** (1.05) 12.89*** (1.11) Candidate List Size 0.21*** (0.02) 0.21*** (0.02) 0.22*** (0.02) Home Candidates 0.04** (0.02) 0.05** (0.02) 0.05** (0.02) Women Candidates 0.005 (0.02) -0.001 (0.02) 0.008 (0.02) Constant 1.02 (12.03) -33.50** (15.87) -13.50 (24.61) Observations 1646 1646 1646 R2 0.5111 0.5134 0.4827 *p < .10 **p < .05 *** p <.01 283 Table 42 - Appendix F - Voting and provincial population size Rent Opportunities and Preference Voting – Population Test #2 (OLS) Model 1 Model 2 Model 3 Estimate (std. err) Estimate (std. err) Estimate (std. err) Civil Service Size (2005) 1.10*** (0.09) Civil Service Size (1999) 1.24*** (0.09) Transfers (by pop) 6.84*** (0.65) Provincial Population (logged) 0.48 (0.35) 1.19*** (0.42) -0.13 (0.39) District Magnitude 0.68*** (0.16) 0.43** (0.16) 0.79*** (0.17) Urbanization 0.09*** (0.02) 0.06*** (0.02) 0.13*** (0.02) Poverty -0.06** (0.02) -0.05** (0.02) 0.09*** (0.02) Ethnic Fractionalization 5.97*** (1.01) 8.02*** (1.07) 12.75*** (1.25) Candidate List Size 0.21*** (0.02) 0.21*** (0.02) 0.22*** (0.02) Home Candidates 0.04** (0.02) 0.05*** (0.02) 0.05** (0.02) Women Candidates 0.005 (0.02) 0.0005 (0.02) 0.008 (0.02) Constant 3.44 (7.01) -7.45 (7.93) -27.71** (11.55) Observations 1646 1646 1646 R2 0.5114 0.5140 0.4825 *p < .10 **p < .05 *** p <.01 Section 5 Indonesia has a long history of separatist conflict. In order to assuage anger and reward local allies, the Indonesian state tends to increase direct transfers to separatist areas. Voters in separatist areas may also be particularly sensitive in their electoral decisions. It is plausible that high numbers of voters in these areas could seek to register a protest against the Indonesian political establishment by supporting individual candidates. Thus separatism could drive both high transfers and preference voting. 284 To test for the possibility that my results are driven by separatist dynamics I simply dropped from the sample electoral districts in provinces with an active separatist movement. Though there is some controversy as to which violent conflicts in Indonesia are ‘separatist,’ I included in the category all electoral districts in Aceh and on the island of Papua. These are the only areas that have had a recognized and active separatist organization for at least a decade. In total, I dropped four electoral districts from the sample. Results appear below. Dropping the districts did no substantively alter the result. Table 43 - Appendix F – Preference voting and separatism Rent Opportunities and Preference Voting – Separatists Excluded (OLS) Model 1 Model 2 Model 3 Estimate (std. err) Estimate (std. err) Estimate (std. err) Civil Service Size (2005) 1.04*** (0.09) Civil Service Size (1999) 1.28*** (0.11) Transfers (by pop) 6.56*** (0.82) District Magnitude 0.51*** (0.16) 0.51*** (0.15) 0.51** (0.19) Urbanization 0.07*** (0.02) 0.08*** (0.02) 0.09*** (0.02) Poverty -0.14*** (0.03) -0.03 (0.03) -0.02 (0.03) Ethnic Fractionalization 4.93*** (0.96) 5.54*** (1.03) 12.31*** (1.11) Candidate List Size 0.21*** (0.02) 0.21*** (0.02) 0.22*** (0.02) Home Candidates 0.04* (0.0) 0.04* (0.02) 0.05** (0.02) Women Candidates -0.005 (0.02) -0.003 (0.02) -0.002 (0.02) Constant 16.66*** (3.22) 10.89*** (3.19) -21.34*** (7.21) Observations 1554 1554 1554 R2 0.4926 0.4867 0.4617 *p < .10 **p < .05 *** p <.01 285 Section 6 The questionnaire and results were provided by the Centre for Strategic and International Studies (CSIS). Ethnic Preferences 97 Dari sejumlah kriteria berikut ini, anda lebih suka memilih presiden di tahun 2009 yang: 1.Ya 2.Tidak 3.Tidak tahu 2. Bersuku bangsa sama dengan anda 1 2 3 97 [From the criteria listed here, I prefer to choose a president in 2009 which:] 1.Yes 2.No 3.Do Not Know 3. [Is the same ethnic group as myself] 1 2 3 Gender 5 Jenis Kelamin responden: 1. Laki-Laki 2. Perempuan 5 Repondent’s Sex: 1. Male 2. Female Coding: 0-1 0. Female 1. Male Education 2 Jenjang dan jenis pendidikan tertinggi yang pernah anda tamatkan: 1. Tidak tamat SD 2. SD 3. Madrasah Ibtidaiyah 4. SD keagamaan lain 5. SMP Umum/Kejuruan 6. Madrasah Tsanawiyah 7. SMP keagamaan lain 8. SMA Umum/Kejuruan 9. Madrasah Aliyah 10. SMA keagamaan lain 11. Perguruan tinggi umum/kejuruan/Diploma 12. Perguruan tinggi agama islam 13. Perguruan tinggi agama lain 286 2 [“Indicate the highest level of education you once completed”] 1. Did not complete elementary school 2. Elementary school 3. Islamic elementary school 266 4. Other religious elementary school 5. Junior high school 6. Islamic junior high school 7. Other religious junior high school 8. High school 9. Muslim high school 10. Other religious high school 11. Post-secondary school 12. Islamic post secondary school 13. Other religious secondary school Coding: 1-5 Scale 1. Did not complete elementary school (1) 2. Completed elementary school (2-4) 3. Completed junior high school (5-7) 4. Completed high school (8-10) 5. Post-secondary school (11-13) Age 6 Usia Responden : ........................................ 6 Respondent’s age: : ........................................ Coding: 1-3 Scale 1. Below 26 2. 26-40 3. 41 above Islam 8 Agama responden: 1. Islam 5. Budha 2. Protestan 6. Konghucu 3. Katolik 7. Kepercayaan 4. Hindu 8. Lainnya...................... 266 Indonesia’s Religious Affairs Department runs a state-sanctioned Islamic educational system. These generally follow the same curriculum used in non-religious schools, supplemented with Muslim content. Non-state religious actors also run independent schools. 287 8 Respondent’s religion: 1. Islam 5. Budhist 2. Protestant 6. Confusionism 3. Catholic 7. Javanese Mysticism 4. Hindu 8. Other...................... Coding: 0-1 0. Non-Muslim 1. Muslim Javanese 9 Suku Bangsa.................................. 9 Ethnic Group.................................. Coding: 0-1 0. Non-Javanese 1. Javanese Urbanization Variable constructed by CSIS using standard classification from the Central Bureau of Statisticis. Coding: 0-1 0. Village 1. Urban 288 Appendix G – Supplement to Chapter 7 Section 1 Predicted values that appear in the text were derived from simulations of Model 1 and Model 2. Results from an interactive model are presented to establish the absence of an interactive effect. Table 44 – Appendix G – Ethnic diversity of party system size [2004-2009] Determinants of Party System Size – National Legislature (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 2004 Model 2 2009 Model 3 2004 Model 4 2009 Ethnic Fractionalization 3.32*** (0.61) 2.03*** (0.51) 9.12** (4.06) 5.60 (3.43) District Magnitude -0.001 (0.60) -0.15 (0.53) 1.88 (1.43) 1.03 (1.24) Urbanization 0.79 (0.94) -4.18*** (0.70) 0.81 0.94) -4.25*** (0.71) Aceh -4.51*** (1.02) -4.48*** (1.02) Ethnic Fractionalization X District Magnitude -2.86 (1.98) -1.81 (1.72) Const 5.19*** (1.32) 9.68*** (1.11) 1.30 (3.00) 7.34*** (2.49) Observations 69 77 69 77 R2 0.3260 0.4747 0.3472 0.4827 *p < .10. **p < .05. *** p <.01. 289 Section 2 Table 45 - Appendix G – Rents and party system size [2004] Determinants of Party System Size, 2004 – National Legislature(OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 – Full Sample Model 2 – Full Sample Model 3 – Homogenous Sample Civil Service Size 0.16*** (0.05) 0.009 (0.06) 0.12 (0.10) District Magnitude 0.37 (0.73) 0.04 (0.67) 2.23** (0.96) Urbanization 2.28** (1.12) 0.87 (1.08) 1.96 (1.32) Ethnic Fractionalization 3.23*** (0.85) Const 3.92** (1.82) 5.03*** (1.68) -0.21 (2.41) Observations 69 69 23 R2 0.1747 0.3263 0.3100 *p < .10. **p < .05. *** p <.01. Section 3 I test the robustness of the results in Table 15 using alternative specifications of rent opportunities. The dependent variable and control variables remain the same as presented. Alternative rent opportunities measures and results appear below: 1. Evidence from Chapter 3 demonstrates that indicators of rent opportunities frequently correlate with provincial rather than municipal ethnic fractionalization. This could be a consequence of governance legacies. Prior to decentralization, the provincial government powerful and competition for provincial power captured attention. Given that provincial and municipal ethnic fractionalization scores frequently diverse, provincial fractionalization is a workable, albeit flawed, proxy for rent opportunities. 2. Chapter 4 demonstrates that rent opportunities induce higher levels of candidate entry. Thus the number of candidates per seat is another potential proxy for indicator opportunities. 3. The 2009 test also includes a district-level measure of rents derived from KPPOD infrastructure scores. 290 Table 46 - Appendix G – Alternative specification of rents [2004] Determinants of Party System Size – National Legislature 2004 (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 Model 2 Model 3 Model 4 Candidates-per-seat 0.58*** (0.12) 0.37*** (0.12) Provincial Ethnic Fractionalization 3.66*** (0.70) 1.39 (1.72) Ethnic fractionalization 2.44*** (0.65) 2.20 (1.52) District Magnitude 2.79*** (0.95) -0.26 (0.60) 2.05** (0.89) -0.07 (0.61) Urbanization -1.98* (1.15) 1.20 (0.96) -1.03 (1.08) 0.94 (0.97) Const -6.27* (3.26) 5.18*** (1.35) -3.23 (3.08) 5.06*** (1.34) obs 69 69 69 69 R2 0.2786 0.3109 0.4085 0.3328 *p < .10. **p < .05. *** p <.01. Table 47 - Appendix G - Alternative specification of rents Determinants of Party System Size – National Legislature 2009 (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Candidates-per-seat 0.09 (0.07) -0.12 (0.08) Provincial Ethnic Fractionalization 1.83*** (0.59) -1.01 (1.27) Infrastructure Scores 0.05** 0.02 0.001 0.03 Ethnic fractionalization 2.63*** (0.65) 2.83** (1.13) 2.02** 0.98 District Magnitude -0.07 (0.75) -0.36 (0.54) -0.41 0.87 -0.84 (0.71) -0.13 (0.53) -0.29 0.85 Urbanization -4.53*** (0.90) -3.84*** (0.73) -3.38*** 1.03 -3.47*** (0.86) -4.33*** (0.73) -3.72*** 1.01 Aceh -4.18*** (1.11) -4.36*** (1.06) -4.50*** (1.01) -4.52*** (1.02) Const 8.67*** (2.48) 9.92*** (1.17) 8.94*** 2.06 12.88*** (2.49) 9.86*** (1.13) 9.79*** 2.04 obs 77 77 51 77 77 51 R2 0.3735 0.4331 0.2717 0.4895 0.4794 0.3332 *p < .10. **p < .05. *** p <.01. 291 Section 4 Party system fragmentation and minor party voting are tightly correlated. Below I present three figures demonstrating the correlation between Major/Minor party support and the effective number of electoral parties. I examined the relationship using three distinct sets of observations:  National Level Party System Size by Percentage Vote for Major Parties in 2004  National Level Party System Size by Percentage Vote for Major Parties in 2009  Municipal Level Party System Size by Percentage Vote for Major Parties in 2004 Figure 35 - Appendix G – Minor party support and electoral fragmentation [2004] 292 Figure 36 - Appendix G - Minor party support and electoral fragmentation [2009] Figure 37 - Appendix G - Minor party support and municipal fragmentation 293 Section 5 Straight-ticket voting in Indonesia is the norm. Parties typically get the same percentage of sub-national votes as national votes. Figure 38 establishes the existence of straight- ticket voting. The dataset contains results from every electoral district in all of Indonesia’s municipal governing bodies. Municipal results were aggregated to match the breakdown of the national-level electoral districts for all 24 parties in all districts outside of Jakarta. Below I plot all 1608 observations. On the X-axis is the percentage of total votes the party received at the municipal level. On the Y-Axis is the percentage of total votes the party received at the national level. The extremely tight relationship indicates that Indonesians most likely support the same party at multiple levels of governance. Figure 38 - Appendix G – Straight ticket voting 294 Section 6 The Table below accomplishes two tasks. First, I show that including the municipalities dominated by the ethnic group “other” does not significantly alter results. Second, I provide the results used to generate Figure 26. Table 48 - Appendix G – Determinants of municipal party system size Determinants of Party System Size – Municipal Legislature (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Variables Model 1 – Full Sample Model 1 – Interactive Civil Service Size .03*** (.01) 0.10*** (0.01) Ethnic Fractionalization 2.43*** (.20) Effective Number of Ethnic Groups 0.03 (0.24) District Magnitude .16 (.20) -0.06 (0.37) Ethnic X Magnitude 0.08 (0.12) Kota .75*** (.15) 0.97*** (0.16) const 5.15*** (.43) 5.39*** (0.77) Observation 1738 1550 R2 0.1613 0.1317 *p < .10. **p < .05. *** p <.01. 295 Section 7 In Table 16, rent opportunities were operationalized through a measure of civil service size. This section presents two alternative measures. Additionally, Model 1 demonstrates that dropping the ethnic “other” dominated municipalities do not significantly affect the results. Table 49 - Appendix G – Alternative specifications of rents (municipal) Determinants of Party System Size – Municipal Legislature (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Variables Model 2 – Alternative 1 Model 3 – Alternative 2 Model 4 – Alternative 1 Homogenous Model 5 – Alternative 2 Homogenous Civil Service Size Provincial Fractionalization 2.51*** (.29) 1.43*** (.29) Candidates per seat .16*** (.02) .23*** (.04) Ethnic Fractionalization 1.28*** (.29) 2.71*** (.20) District Magnitude .23 (.21) .12 (.21) -.25 (.34) -.46 (.33) Kota .80*** (.15) .47*** (.15) 1.25*** (.25) .77*** (.25) const 4.47*** (.45) 3.34*** (.55) 5.73*** (.72) 3.68*** (.85) Observation 1550 1550 630 630 R2 0.2054 0.1885 0.0661 0.0800 *p < .10. **p < .05. *** p <.01. 296 Section 8 This section provides additional results from the testing of Hypothesis 4. Dependent variable: Effective Number of Electoral Candidates [ENEC] Independent variables: 1. Ethnic fractionalization (municipal) – 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous. Measured at the municipal level 2. Urbanization – 0-1 measure, 0 being no 0% of citizens living in areas designated ‘kota’ by central government, 1 being 100% of citizens living in areas designated ‘kota.’ 3. Candidates – Number of candidates competing (logged) 4. Year 2009 – Dummy if 2009 election Table 50 - Appendix G – Determinants of upper house fragmentation Determinants of Party System Size – Upper House (DPD) Election (OLS Regression) Dependent Variable: Effective Number of Electoral Candidates Model 1 – Full Sample Model 2 – 2004 Sample Model 3 – 2009 Sample Ethnic Fractionalization 2.03 (1.86) 3.53 (2.62) .15 (2.81) Number of Candidates (logged) 10.33*** (1.19) 9.86*** (1.62) 11.13*** (1.91) Urbanization -7.54** (2.97) -3.91 (3.40) -10.78** (4.46) Year 2009 -.05 (1.06) Const -16.77*** (4.44) -17.59*** (6.28) -17.12** (6.94) Observation 65 32 33 R2 0.5768 0.5768 0.5888 *p < .10. **p < .05. *** p <.01. 297 Ethnic Fractionalization by Effective Number of Electoral Candidates Figure 39 - Appendix G – Upper house fragmentation and candidate numbers 298 Appendix H – Supplement to Chapter 8 Section 1 - District-Level Party System Size in 1999 Dependent Variable:  Effective Number of Electoral Parties in the National District Independent Variables:  Ethnic fractionalization (municipal) – 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous.  Urbanization – 0-1 measure, 0 being no 0% of citizens living in areas designated  District Magnitude – Number of Seats in the District (logged) Table 51 - Appendix H – Determinants of party system size [1999] Determinants of Party System Size – National Legislature (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 Model 2 Ethnic Fractionalization -0.25 (0.89) -0.24 (1.02) District Magnitude -0.07 (0.36) Urbanization 0.52 (1.62) Const 4.55 (0.64) 4.54 (1.44) Observations 26 26 R2 0.0032 0.0087 *p < .10. **p < .05. *** p <.01. 299 Section 2 - Municipal-Level Party System Size in 1999 Dependent Variable:  Effective Number of Electoral Parties, Municipal-Level Independent Variables:  Civil Service Size – Provincial % of modern sector workers employed in civil service  Ethnic Fractionalization - 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous.  Effective Number of Ethnic Groups – Alternative diversity measure constructed using the same process as the Effective Number of Electoral Parties. In practial terms, 1 / (1 - Ethnic Fractionalization)  City – 1 if ‘kota’, 0 if ‘kabupaten’  District Magnitude – Number of eats in the national district (logged)  Civil Service X Ethnic Groups – Interactive variable constructed by multiplying Effective Number of Ethnic Groups and Civil Service Size Table 52 - Appendix H – Determinants of municipal party system size [1999] Determinants of Party System Size – National Legislature (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 Model 2 Model 3 Model 4 Civil Service Size -0.02** (0.01) -0.06** (0.03) Ethnic Fractionalization 0.48** (0.22) Effective Number of Ethnic Groups 0.11 (0.10) District Magnitude (logged) -0.32*** (0.12) City 0.25 (0.19) Civil Service X Ethnic Groups -0.005 (0.006) Const 4.29*** (0.15) 3.84 (0.12) 5.51*** (0.67) Observations 312 311 285 R2 0.0111 0.0153 0.0740 *p < .10. **p < .05. *** p <.01. 300 Figure 40 - Appendix H – Marginal effect of diversity on electoral fragmentation Section 3 - Party System Change over Time Comparing 1999 and 2009 Note: Technically there were only 26 national electoral districts in 1999.267 To compare over time I reconstructed the 1999 results to correspond to the 2009 electoral districts. All references to “Effective Number of Parties – 1999” refer to the reconstructed results. X-Axis: Effective Number of Parties – 1999 Y-Axis: Effective Number of Parties - 2009 267 Excluding East Timor. 301 Figure 41 - Appendix H – Party system size 1999-2009 Growth over Time Note: Growth in Effective Number of Electoral Parties = ENEP 2009 – ENEP 1999 X-Axis: Growth in Effective Number of Electoral Parties Y-Axis: Effective Number of Parties - 2009 Figure 42 - Appendix H - Party system expansion 302 Section 4 - Determinants of Party System Expansion Dependent Variable:  Expansion of Effective Number of Electoral Parties in the National District Independent Variables  Civil Service Size – Provincial % of modern sector workers employed in civil service  Ethnic fractionalization – 0-1 measure, 0 being completely homogenous, 1 being completely heterogonous  Urbanization – 0-1 measure, 0 being no 0% of citizens living in areas designated ‘kota’ by central government, 1 being 100% of citizens living in areas designated ‘kota.’  Aceh – 1 if the district is in the province of Aceh, 0 if outside. Table 53 - Appendix H – Determinants of party system expansion Determinants of Party System Size – National Legislature (OLS Regression) Dependent Variable: Effective Number of Electoral Parties Model 1 Model 2 Civil Service Size 0.11*** (0.03) Ethnic Fractionalization 1.82*** (0.54) Urbanization -3.97*** (0.83) -4.94*** (0.78) Aceh -6.86*** (1.16) -6.46*** (1.14) Const 5.08*** (0.54) 5.60*** (0.43) Observations 77 77 R2 0.4986 0.5016 *p < .10. **p < .05. *** p <.01. 303 Section 5 - Party change and party system expansion Dependent Variable:  Expansion of Effective Number of Electoral Parties in the National District Independent Variable:  Change in % of Votes for each Major Party o Example: West Kalimantan District  Golkar Losses = Golkar Vote 1999 [29%] - Golkar Vote 2009 [14%] Golkar Losses = 15% This process was repeated for each of the ‘Big-5’ parties of 1999. Table 54 - Appendix H – Party change and party expansion Determinants of Party System Expansion, 1999-2009 – National Legislature (OLS Regression) Dependent Variable: Change in Effective Number of Electoral Parties Model 1 - Golkar Model 2 – PDIP Model 3 - PKB Model 4 – PPP Model 5 - PAN Party Losses 0.77*** (0.18) -0.41 (0.27) 0.26 (0.27) -0.77 (0.49) -0.25*** (0.32) Const 3.53*** (0.29) 5.05*** (0.56) 4.16*** (0.29) 4.72*** (0.36) 4.59*** (0.19) Observations 77 77 77 77 77 R2 0.1941 0.0287 0.0124 0.0317 0.4351 *p < .10. **p < .05. *** p <.01. 304 Section 6 - Golkar’s Decline Over Time Dependent Variable:  % of Electorate Vote for Golkar Independent Variables:  Civil Service Size – Provincial % of modern sector workers employed in civil service  Election Number – Election Number since fall of Suharto (1999=1, 2004=2, 2009=3)  Civil Service X Election – Civil Service Size X Election Number  Poverty – % of population designated as living in poverty  Leader Home – Dummy variable capturing whether Golkar leader from the ‘Iramasuka’ region. If the leader is from ‘Iramasuka’, all provinces within the region are valued ‘1’, if not the provinces are valued ‘0.’ Table 55 - Appendix H – Determinants of Golkar support Determinants of Golkar Support – National Legislature (OLS Regression) Dependent Variable: % Golkar Vote Model 1 - 1999 Model 2 - 2004 Model 3 - 2009 Model 4 - Pooled Model 5 - Pooled Civil Service Size 0.42 (0.35) 0.56* (0.31) 0.20 (0.18) 0.39* (0.21) 1.41* (0.74) Election Number -6.50*** (1.40) -0.51 (3.13) Civil Service X Election -0.49* (0.28) Leader Home 32.23*** (4.48) 5.18** (2.35) 16.73*** (2.42) 16.52*** (2.46) Poverty 0.31* (0.16) -0.03 (0.17) -0.04 (0.12) 0.12 (0.14) 0.10 (0.14) Const 12.29*** (4.11) 18.41*** (4.06) 13.71*** (2.23) 27.25*** (4.62) 15.31*** (8.56) Observations 26 26 26 78 78 R2 0.7706 0.1581 0.2620 0.5172 0.5539 *p < .10. **p < .05. *** p <.01. 305 Appendix I – List of interviews This appendix contains all cited interviews, organized by date. This list does not include informal discussions with scholars, government workers, and NGO activists. All interviews took place between January and June 2009. This was the period directly before and after the legislative election. Interviews were solicited in three provinces: Jakarta, North Sumatra, and West Nusa Tenggara. The latter two provinces were initially selected to provide variation in ethnic structure: North Sumatra is a relatively diverse province whereas West Nusa Tenggara is relatively homogenous. Interviews in Jakarta targeted the major party’s central offices. All interview participants signed a consent form making them aware of the project subject and the author’s affiliation. Interviews were structured around a set of base questions that varied depending on whether the interviewee was a candidate / local party activist or a figure within the party’s national office. Interviews were conducted in Bahasa Indonesia with the aid of an interpreter. To maintain anonymity I refer to respondents only by their broad titles. Table 56 - Appendix I – List of interviews Date Location Interviewee February 25, 2009 Medan, North Sumatra Former PDI-P DPRDII candidate (Central Tapanulli) February 27, 2009 Kabanjahe, North Sumatra Gerindra DPRD II candidate (Karo) February 27, 2009 Kabanjahe, North Sumatra Partai Pengusaha dan Perkerja Indonesia DPRD II candidate (Karo) February 27, 2009 Kabanjahe, North Sumatra Partai Karya Perjuangan Regional Chairman (Karo) February 28, 2009 Kabanjahe, North Sumatra Partai Kasih Demokrasi Indonesia, DPRDII candidate (Karo) February 28, 2009 Kabanjahe, North Sumatra Partai Keadilan dan Persatuan Indonesia Team Success leader for DPRDII candidate (Karo) February 28, 2009 Kabanjahe, North Sumatra Partai Demokrat DPRDII candidate (Karo) March 4, 2009 Medan, North Sumatra Partai Demokrat DPRD candidate March 7, 2009 Pematang Siantar, North Sumatra Hanura DPRD candidate March 7, 2009 Pematang Siantar, North Sumatra Partai Merdeka DPRDII candidate (Simalungun) March 7, 2009 Pematang Siantar, North Sumatra Barisan Nasional DPRDII candidate (Pematang Siantar) March 11, 2009 Medan, North Sumatra PPP DPR-RI candidate April 6, 2009 Bengkel, West Nusa Tenggara Hanura DPRDII candidate (Lombok Barat) April 8, 2009 Bengkel, West Nusa Tenggara Golkar DPRDII candidate (West Lombok) April 8, 2009 Bengkel, West Nusa Tenggara DPD West Nusa Tenggara candidate 5 June, 2009 Jakarta Golkar national office official 306 Date Location Interviewee 8 June, 2009 Jakarta PKS national party staffer 10 June, 2009 Jakarta PAN national office official 11 June, 2009 Jakarta PPP national office official 15 June, 2009 Jakarta PDI-P national office official 18 June, 2009 Jakarta PD national office official