The Political Economy of Clientelism: A Comparative Study of Indonesia’s Patronage Democracy

Abstract

What kind of economic development curtails clientelistic politics? Most of the literature addressing this relationship focuses narrowly on vote buying, resulting in theories that emphasize the importance of declining poverty rates and a growing middle class. This article employs a combination of ethnographic fieldwork and an expert survey to engage in a first-ever, more comprehensive comparative study of within-country variation of clientelistic politics. I find a pattern that poorly matches these dominant theories: Clientelism is perceived to be less intense in rural, poverty-prone Java, while scores are high in relatively wealthy yet state-dependent provincial capitals. On the basis of these findings, I develop an alternative perspective on the relationship between economic development and clientelism. Emphasizing the importance of societal constraints, I argue that the concentration of control over economic activities fosters clientelism because it stifles the public sphere and inhibits effective scrutiny and disciplining of politico-business elites.

Introduction

Political clientelism refers to the practice of providing personal favors—jobs, contracts, welfare support, money, and so forth—in exchange for electoral support. Long considered a premodern artifact destined to be swept away by modernization, studies from Asia (Chandra, 2004) and Africa (van de Walle, 2007) to South America (Auyero, 2001) to Eastern Europe (Kopecký & Spirova, 2011) suggest that democratization is making clientelistic politics less exploitative but not necessarily less pervasive. As the recent World Development Report (The World Bank, 2017) attests, there is growing recognition that this mode of electoral mobilization underlies a host of developmental challenges—from corruption and inadequate public service provision to ethnic violence and a weak rule of law (see also Fukuyama, 2011; Mungiu-Pippidi, 2015).

Yet despite its pervasiveness and impact, the study of the conditions favoring or curtailing political clientelism “is, perhaps surprisingly, still in its beginnings” (Roniger, 2004, p. 369; see also Hicken, 2011) as there are “surprisingly few systematic comparative studies on clientelism” (Kitschelt & Wilkinson, 2007, p. 3, see also Kitschelt, 2000). This is largely due to a key methodological challenge: Despite some recent progress, the secretive and shadowy nature of clientelistic practices has hampered attempts to assess and compare in a reliable fashion the extent to which political systems are clientelistic. This methodological challenge has contributed to a relative neglect of the comparative study of subnational democratization (Gibson, 2005; Sidel, 2014) as it has remained difficult to assess whether, and to what extent, the character of local politics differs across regions.

Combining an expert survey with ethnographic fieldwork on election campaigns, this article applies a new research method to engage in such a comparative study of subnational politics. Building on a similar initiative aimed at national comparisons pioneered by Herbert Kitschelt (Kitschelt, Freeze, Kolev, & Wang, 2009; Kitschelt & Kselman, 2013), this study uses assessments of local experts to gauge and compare the intensity of clientelistic practices in 38 districts across Indonesia. I find considerable yet consistent variation across Indonesia: While clientelistic exchanges are perceived to be less intense throughout Java, particularly the cities, experts consider these exchanges to be much more pervasive in eastern Indonesia, including its provincial capitals.

On the basis of these survey results and my fieldwork on election campaigns in Indonesia, I develop an alternative perspective on the relation between economic development and clientelistic politics. Most of the literature assessing this link adopts a relatively narrow focus on vote buying, resulting in theories that center on the ways in which economic development affects the costs and yields of clientelistic strategies for politicians and voters. This literature expects clientelistic politics to be pervasive when poverty rates are high, the middle class is small, and urbanization limited. The results of the expert survey match these expectations rather poorly.

Deriving inspiration from the growing literature on subnational democratization, I use my fieldwork to develop an alternative “constraint perspective” that focuses on how economic development can foster the emergence of social forces capable and willing to expose and criticize clientelistic practices. In this view, not only the degree but also the character of economic development matters: Economic development can curtail clientelistic practices when it supports the emergence of a more autonomous civil society and a more open public sphere. This approach differs not only in terms of the mechanisms it proposes for linking economic development and clientelistic politics, but also in terms of its predictions: This perspective expects clientelistic politics to be most pervasive when local economies are not diversified and highly state-dependent. These hypotheses are borne out by the results of the expert survey.

The article proceeds as follows. The section “What Drives Clientelistic Politics?” briefly reviews available economic explanations for the pervasiveness of clientelistic practices. The sections “Indonesia’s Patronage Democracy” and “Election Campaigns and Local Economies” provide a brief introduction to Indonesia’s patronage democracy and further develops these economic explanations using fieldwork material. The sections “The Comparative Study of Clientelistic Politics” and “The Clientelism Perception Index” discuss the expert survey and the construction of the “clientelism perception index.” The section “An Overview of the Findings” presents the findings while the section “What Makes Java Less Clientelistic?” assesses how well these different explanations can account for the patterns found. The final section concludes.

What Drives Clientelistic Politics?

Despite the relative paucity of empirical analysis of conditions favoring political clientelism, the literature on the topic contains a rich set of explanations for the pervasiveness of clientelistic electoral strategies. Some of these explanations—such as the character of electoral institutions or the relative youthfulness of a democracy (Keefer & Vlaicu, 2008)—are useful for explaining differences between countries, but cannot offer much leverage when accounting for subnational variation. In this article, I will focus on economic factors as explanations for subnational variation. In the rapidly expanding literature on clientelism, arguably the most widely shared conclusion is that economic development—defined mostly as growth of per capita income—undermines and curtails clientelistic politics. This conclusion is supported by a considerable body of (survey-based) research that shows that poor countries are more likely to be clientelistic (Bustikova & Corduneanu-Huci, 2009; Keefer, 2007; Kitschelt & Wilkinson, 2007), that poor voters are more likely to receive and respond to money or other clientelistic incentives (Brusco, Nazareno, & Stokes, 2004; Stokes, 2005; Stokes, Dunning, Nazareno, & Brusco, 2013, pp. 152-171), and that politics in poor regions is more likely to take a clientelistic form (Remmer, 2007; Wantchekon, 2003). Some argue that the relationship is curvilinear with clientelistic politics first intensifying at low levels of development (Kitschelt & Kselman, 2013).

What is much less well established, however, is why and how economic development curtails clientelistic practices. Two important reviews of the literature both observe that, “[while] the affinity between poverty (inequality) and clientelism is a settled fact, . . . the mechanisms linking the two, and the direction of causality, are not” (Hicken, 2011; Stokes, 2007, p. 623). The literature on clientelism generally analyses this link between economic development in terms of how rising incomes and the attendant growth of the middle class reduces the yields and increases the costs of clientelistic strategies for politicians (see Stokes, 2007). In the words of Magaloni, Diaz-Cayeros, and Estévez (2007), “As a country develops and the pivotal voter becomes wealthier, clientelism should erode as a dominant form of political exchange simply because it becomes too costly” (p. 203). As shorthand, I will dub this approach the “cost perspective.” Along these lines, some argue that as poor voters are more risk-averse and, hence, more easily tempted by clientelistic benefits (Brusco et al., 2004; Kitschelt, 2000; Scott, 1972a), rising incomes reduce the marginal utility of personal inducements for voters, thus increasing the overall cost of clientelistic practices (Calvo & Murillo, 2004; Dixit & Londregan, 1996; Magaloni et al., 2007; Robinson & Verdier, 2013). A second set of arguments focuses on how economic development generates a bigger middle class that might dislike clientelistic practices (Kitschelt & Wilkinson, 2007; Weitz-Shapiro, 2014). A third group focuses on how economic development complicates vote monitoring: In the less tightly knit cities, clientelistic exchanges are more difficult to enforce and thus less effective (Lehoucq, 2007; Stokes et al., 2013). A limitation of this literature is that it focuses mostly on vote buying (and, sometimes, access to welfare programs) while clientelistic politics also takes other forms, such as the conditional provision—and withholding—of government jobs, contracts, or public services (Mares & Young, 2016). There have, to my knowledge, so far been no attempts to assess whether abovementioned arguments could be extended to these other dimensions of clientelistic politics.

An alternative perspective could be derived from the adjacent literature on subnational democratization. The object of research of these studies is not clientelism—they focus on “subnational authoritarianism” (Gibson, 2005), “hybrid democratic regimes” (McMann, 2006), and the “unevenness of democratization” (Behrend, 2011). Yet a pertinent and useful conclusion of these studies is that they relate subnational variation in the character of local politics to a concentration of economic control—that is, the degree to which control over important economic activities is concentrated in the hands of a small elite. Kelly McMann, for example, compared two districts (oblasti) in Russia and two in Kyrgyzstan. She concluded that differences in the character of local politics can be attributed to varying levels of “economic autonomy,” which she defines as “the ability to earn a living independent of the state” (McMann, 2006, p. 20). These studies conclude that politics takes the form of a “closed game” (Behrend, 2011, p. 154) in areas with a “concentration of control over the commanding heights of a local economy in the hands of a single individual, family, clan, clique, or organization” (Sidel, 2014, p. 177). Such analyses resonate with the literature on the political economy of democratization, which shares a focus on the conditions conducive to the development of a state-autonomous civil society (e.g., Rueschemeyer, Stephens, & Stephens, 1992).

From this literature and my fieldwork—discussed below—I derive the hypothesis that the concentration of control over economic activities fosters clientelism because it stifles the public sphere and inhibits effective scrutiny and disciplining of politico-business elites. As shorthand, this approach could be termed the “constraint perspective,” as it focuses on the conditions underlying the capacity and willingness of citizens and business actors to resist, expose, and criticize clientelistic practices. When economic development boosts the economic independence of citizens from ruling elites, a public sphere capable of constraining clientelistic practices can emerge. Concentration of economic control can take different forms—ranging from “company towns” to natural resource dependency and a dependency of government budgets. Again taking a cue from the literature on subnational democratization, I will focus particularly on state dependency as well as limited economic diversification as two key indicators of concentration of economic power. A dependency on state budgets fosters concentrated control because it concentrates power in the hands of ruling elites capable of allocating state resources. When economic diversification is limited—as evidenced by, for example, small industry, trade, and services sectors—economic power is generally more concentrated because of a lack of sources of income outside the ambits of (ruling) elites.

This perspective not only differs from the “cost perspective” in terms of its interpretation of the conditions favoring clientelistic politics. Underlying these perspectives are different ideas about the kinds of mechanisms that curtail clientelistic practices. The “cost perspective” dominant in the clientelism literature largely adopts a rational-choice and individualist perspective, focusing on the factors that affect the cost-benefit analyses of both voters and politicians while largely ignoring questions of (economic) power and the character of the public sphere. In contrast, the constraint perspective argues that it is not economic growth per se, but rather the wider distribution of power and control over economic activities that facilitates a curtailment of clientelistic activities. In this view, politicians do not shift away from clientelistic practices because the yields are decreasing, but rather because of the capacity and willingness of civil society to oppose such practices by, for example, exposing them and propagating a moral discourse that promotes a rule-bound distribution of resources.

A similar contrast involves the role of the middle class and industrialization. The clientelism literature sees the growth of the middle class as conducive to more programmatic politics, based on the assumption that a middle class is more likely to despise clientelistic practices and value the rule of law. As this middle class would be less likely to vote for candidates engaging in clientelistic strategies, the growth of the middle class will thus increase the “audience cost” (in terms of loss of votes) of clientelism (Weitz-Shapiro, 2014) and generate a “constituency for universalism” (Shefter, 1994, p. 28). In contrast, the constraint perspective suggests that it is not the size of the middle class per se, but rather the extent to which this middle class depends on (a privileged access to) the state for its livelihoods. Similarly, in contrast to the cost perspective, industrialization might serve to curtail clientelistic practices not because it generates higher incomes, but rather because it can foster a more even distribution of power and a more autonomous civil society. As control over sources of livelihoods becomes more dispersed beyond a narrow oligarchic elite that controls the state, civil society as well as rival oligarchs becomes stronger and freer to resist and criticize clientelistic practices of power holders. An assessment of the validity of these different perspectives is thus needed not only to identify the kinds of conditions fostering clientelism, but also to understand the mechanisms that generate change: Is it because of rising costs or growing constraints that economic development curtails clientelistic practices?

Comparative assessments of the pervasiveness of clientelistic practices can help to address this question. These two different perspectives come with very different predictions. The cost perspective predicts that clientelistic politics is more pervasive in regions with lower poverty rates, a more rural population, and a bigger middle class. The constraint perspective, in contrast, predicts that clientelistic politics is more pervasive in regions with limited economic diversification and a higher degree of dependence on state budgets. These two sets of economic conditions are correlated to a certain degree—as economic growth often goes hand in hand with economic diversification.¹ Yet the degree of economic development (as measured, for example, in terms of income) can be distinguished from the character of economic development, which concerns the type of economic variety generating economic grown. A growth in income does not necessarily lead to a dispersion of economic control. In Indonesia, for example, the correlation between economic growth and diversification is far from perfect as there is considerable variation between, for example, poor but diversified rural Java and relatively wealthy but state-dependent provincial capitals outside Java. This heterogeneity makes it possible to compare the predictive capacity of these perspectives. I will use a regression analysis to probe the extent to which different economic conditions can account for (expert perceptions of) varying levels of clientelism across Indonesia.

Indonesia’s Patronage Democracy

Indonesia’s new democracy calls out for such a comparative study. After the New Order’s highly centralized, authoritarian, and clientelistic rule ended in 1998, Indonesia’s regions were all subjected to the same combination of democratic reform and decentralization, leading to the adoption of direct elections for regional leaders in 2004. The character of these direct elections for district heads (bupati or, in cities, walikota) and governors are the focus of this article. Some observers consider these elections to be highly clientelistic, dubbing Indonesia a “patronage democracy” (Nordholt & van Klinken, 2007; Simandjuntak, 2009; van Klinken, 2009) run by “predatory elites” (Hadiz, 2010; Winters, 2011). Others have argued that an increasingly vocal civil society is using newly gained civil liberties to curtail clientelistic practices (Mietzner, 2012; Stanley, Priyono, & Törnquist, 2004). Such debates, however, are largely waged in general terms and no comparative assessment has ever been made.

For an interpretation of the results presented in this study, three general features of the character of such local government head elections need to be briefly mentioned. A first important feature is the relative independence of candidates from political parties. Indonesia’s electoral procedures demands that candidates, to be eligible, obtain the support of political parties that, collectively, control at least 20% of the seats in the local parliament.² The result of this requirement is that candidates rarely emerge out of party ranks. A candidature more often rests on a capacity to negotiate (and buy) the support of a coalition of parties (Aspinall, Berenschot, & Hendrawan, 2017).

A second and related aspect is the weak mobilizational capacity of political parties. Suharto had banned local political activity during the New Order, and this particular inheritance continues to limit the capacity of political parties to provide manpower to sustain election campaigns. Instead, candidates build their own campaign networks, referred to as tim sukses (Aspinall, 2014b). To build these campaign networks, candidates need to woo supporters and build connections with local elites, such as religious leaders, businessmen, and local bureaucrats. Candidates often need to (promise to) provide privileged access to state resources—jobs, business licenses, government contracts—to attract such supporters to build their campaign organization. In other words, the relationships that animate campaign organizations are largely clientelistic in nature (see Aspinall & Berenschot, 2018; Aspinall & Sukmajati, 2016).

Third, campaign costs are massive and largely borne by candidates themselves. They need massive funds to buy the required support from political parties, fund campaign activities, and to engage in (often extensive) vote buying (Aspinall, 2014a). These high costs of election campaigns foster corruption and force politicians to engage in various kinds of (clientelistic) deals with business actors (Hadiz, 2010; Mietzner, 2013). The high costs of election campaigns contribute to the oligarchic nature of Indonesian politics as economic elites often succeed in translating material wealth into political power (Winters, 2011). This is visible in the three regions I will focus on below: The current governor of Central Kalimantan is the nephew of a palm oil and timber tycoon while the governor of Lampung is the son of a director of the country’s largest sugar-producing company with large plantations in the province. In these provinces, the capture of political power served to acquire the necessary licenses and permits to expand business opportunities. While the candidate I shadowed in Tangerang also hailed from a wealthy family—they own a chain of hospitals—the economic prosperity of this family seemed less closely dependent on his political success.

Election Campaigns and Local Economies

In 2013 and 2014, I studied the election campaigns in these provinces during 13 months of fieldwork in three very different districts. I followed elections in the industrialized and relatively affluent Tangerang city (a suburb of Jakarta), the rural backwater Lampung, and natural resource–rich Central Kalimantan. Local politics in these regions differed considerably, in ways that illustrate the interplay between the character of the local economy and the capacity of citizens to resist and constrain clientelistic practices. On the basis of this fieldwork, I identify four key mechanisms linking a dispersion of economic power and the capacity of citizens and business owners to constrain clientelistic practices.

A first mechanism concerns the collusive embrace between politics and business that characterizes local politics in many parts of Indonesia. In both Central Kalimantan and Lampung, campaign funding was largely driven by rather crass tit-for-tat deals. State-dependent entrepreneurs involved in, for example, construction or, in Kalimantan, palm oil and mining provided candidates with campaign funds with the expectation of obtaining privileged access to government contracts and concessions for plantations or mines. As most economic activity in these regions is tied to either government budgets or government approval, business actors can hardly sustain themselves without support from power holders. Politicians, from their end, can hardly fund their campaign without engaging in such clientelistic exchanges. The election campaigns I observed in Kalimantan amounted to a contest between different politico-business coalitions.

In contrast, the more diversified economies of Java’s cities contain a more varied range of economic actors, including those who do not rely on government contracts or concessions or those who, producing for the global economy, are harmed by these practices. While the politicians I shadowed in Tangerang were not foreign to under-the-table deals, the availability of such wider sources of campaign funding gave them considerable more leeway when dealing with business actors. Furthermore, the availability of a more varied pool of economic elites discourages politicians from engaging in the kinds of collusive deal-making found in Central Kalimantan because such deals face more scrutiny from rival economic elites (see von Luebke, 2012). As one perceptive study comparing Manado (Sulawesi) and Solo (Java) concluded, “The multi-sectoral nature of the economy [of Solo] renders the emergence of narrowly defined collusive arrangements less probable” (Patunru, McCulloch, & von Luebke, 2012, p. 805). Such observations suggest that when economic power is more dispersed, politicians are not only less dependent on clientelistic exchanges with business actors but they also face greater scrutiny of such exchanges.

A second mechanism concerns the capacity of politicians to enforce clientelistic exchanges. When livelihoods are highly dependent on the state (or on a few elite-controlled businesses), a refusal of citizens (or business owners) to provide electoral support after having received personal favors might lead to more serious repercussions. For example, in both Lampung and Kalimantan, I observed not just civil servants but also their whole families supporting the election campaign of ruling politicians. A failure to provide such support could lead to a demotion or dismissal, which could endanger the prosperity and status of the extended family. Similarly, the recipients of state benefits—from government projects, social assistance funds, or business licenses—would put their livelihoods at risk if they would not reciprocate at election time. For example, both contractors as well as (in Central Kalimantan) managers of palm oil plantations generally attempt to influence the voting behavior of their workers—threatening these workers with dismissal if this voting advice is not followed. These two types of business are highly dependent on state budgets and regulation, making it difficult for business owners (and their workers) to refuse to provide such electoral support. In other words, in areas with concentrated economic control, the economic dependencies are such that citizens are more likely to face serious repercussions when resisting or critiquing clientelistic practices.

A third mechanism concerns the autonomy of local civil society. The presence of a large industrial sector in Tangerang sustained relatively strong labor unions. The unions not only actively scrutinized the behavior of local politicians, but also facilitated a nascent form of programmatic politics by offering electoral support to politicians who promised to raise the minimal wage (see Caraway, Ford, & Nugroho, 2015; Tans, 2012). The activities of these labor unions and their critique of ruling elites has the important side effect of popularizing a moral discourse that generates aversion to a clientelistic distribution of state resources. In Tangerang, I also encountered the curious phenomenon of “bodrex NGOs.” With this term, politicians and bureaucrats referred to (and complained about) small organizations or individuals who threaten to report any kinds of wrongdoings (“bodrex” refers to an all-purpose medicine) to the local press if they are not paid off. While this kind of extortion hardly amounts to civic behavior, the threat of being exposed can serve to constrain corrupt and clientelistic practices.

In contrast, in Lampung and particularly Kalimantan, civil society organizations seemed much more reluctant to expose or criticize power holders. The various (often identity-based) cultural associations, farmer groups, or local vigilante groups I encountered were regularly mocked as being plat merah or “red license plate.” With this reference to the license plates of government cars, my informants lamented the closeness between such associations and the local government. These organizations did not engage in a similar critique of clientelistic practices. Instead, most associational activity serves to curry favor with politicians to obtain government projects. For example, one of Central Kalimantan’s most prominent organizations, the association of Dayak customary leaders (Dewan Adat Dayak [DAD]) now regularly campaigns for ruling politicians after local customary leaders started receiving government stipends since 2010. In a similar fashion, the local ethnic associations (paguyuban) operate basically as vote-pooling devices, using their purported influence over voters as means to extract clientelistic favors. This behavior makes sense in a state-dependent economy: Lacking independent sources of funding outside the ambit of ruling elites, the few remaining critical voices are more easily co-opted or starved of funds and career opportunities. In short, in areas where control over economic activities is relatively concentrated, civil society organizations struggle to obtain independent sources of funding and they are more likely to face negative consequences if they adopt a critical stance.

A fourth and related mechanism concerns the nature of the public sphere. In Kalimantan and Lampung, the incomes of local newspapers are heavily dependent on local governments and politicians because they are by far the largest advertisers. In some of the newspapers in Central Kalimantan, up to one third of the pages are filled with advertisements of government agencies, district heads, and local parliaments. This dependency shapes editorial policies. Journalists recounted to me how their editors refused to publish critical articles out of a fear of losing advertisement revenue. This might also be the reason why I did not find any “bodrex NGOs” in Central Kalimantan: Threats to power holders to expose wrongdoings are less effective when newspapers can be easily persuaded to ignore such stories.

In contrast, the more diversified economy of Tangerang generates a more diverse set of advertisers, making a fall out with power holders less damaging for a newspaper. This widens the scope for critical reporting. The resulting openness of the public sphere could serve to criticize and curtail clientelistic practices, sometimes in surprisingly direct ways: One of my informants, the head of the campaign team of Tangerang’s sitting mayor, lost out on a plum patronage post as director of a state-owned firm after an article appeared in a local newspaper in which a local nongovernmental organization (NGO) criticized the mayor for engaging in “collusive” appointments.³ This cooperation between a critical NGO and an independent newspaper forced the local mayor to abort a clientelistic appointment.

These unavoidably sketchy comparisons⁴ between politics in Lampung, Central Kalimantan, and Tangerang substantiate the constraint perspective, in the sense that these comparisons yielded four key mechanisms linking the concentration of economic control to the limited capacity and willingness of local business actors, civic associations, and newspapers to resist and criticize clientelistic practices. We will now turn to the expert survey to assess to what extent this perspective can indeed account for variation across Indonesia.

The Comparative Study of Clientelistic Politics

While our understanding of clientelism has benefited greatly from qualitative, ethnographic studies on clientelistic politics in different settings, such studies lack methodological and conceptual tools to compare regions in terms of their relative intensity of clientelistic practices. Since the early 2000s, a new wave of research has adopted a more comparative approach using either survey data (e.g., Brusco et al., 2004; Stokes, 2005) on or various statistical proxies—such as perceptions of corruption and rule of law measures (Keefer, 2007; Keefer & Vlaicu, 2008), the rate of government spending on personnel (Magaloni et al., 2007), or mayoral discretion in the spending of welfare budgets (Weitz-Shapiro, 2014). These methods have various drawbacks. General voter surveys can only address one dimension of clientelistic practices. Surveys can only ask questions about vote buying and thus ignore other forms of clientelistic exchanges such as the provision of jobs, public services, or government contracts. The drawback of using proxies is not just that the data are not always very reliable, but also that the link between the adopted statistical measure and clientelism is often questionable. For example, it cannot always be assumed that more spending on government personnel or a weak rule of law is an indication of a higher intensity of clientelistic vote mobilization. As Muno’s (2010) and Hicken’s (2011) overviews of different methods conclude, compared with these alternatives, an expert survey—particularly if executed alongside ethnographic fieldwork—can be cheaper, more comprehensive, and more reliable. An expert survey can pay attention to a range of clientelistic goods—from government jobs and contracts to public services and welfare support. Furthermore, compared with those relatively distant statistical proxies mentioned above, the perceived intensity of clientelistic practices offers a more direct measure.

An expert survey also has its drawbacks. Expert assessments can still be considered a proxy as they provide perceptions of how common clientelistic practices are. The relation between perceptions and the actual reality of clientelistic practices is a complex one. On one hand, one could advance various reasons—partisan bias, lack of information, awareness—that could lead to an over- or underestimation of clientelistic practices. On the other hand, it can be argued that the perceived reality of clientelistic practices is just as important as its “actual” reality. The perception that the distribution of state benefits is contingent on electoral support is more than a proxy for the “reality” of clientelistic practices because it is this perception—rather than “reality”—that motivates people to engage in clientelistic relationships. These perceptions are part of that reality.

The use of an expert survey to study clientelistic practices faces two main methodological challenges. The first concerns the risk that the partisan bias of the experts might influence their assessments. Political observers who are supportive of ruling politicians might be inclined to downplay the incidence of clientelistic practices, while dissidents might provide overestimations.⁵ Second, as with all surveys, confusion about questions and answer categories can reduce reliability, requiring close attention to both the wording of the survey and its execution.

The expert survey presented in this study was successfully⁶ executed in 38 districts in 16 Indonesian provinces. For each district, 14 experts were interviewed. In total, 533 interviews were successfully concluded. The survey was executed between April and June 2014. To minimize these two possible sources of distortion, the following efforts were made in the implementation of the survey. I decided to involve only locally rooted researchers with considerable experience. With the help of two national research institutions, 35 researchers were recruited who had previously collaborated on other large nationwide survey projects.⁷ To ensure the quality of the survey, all these researchers were trained and the interviews were recorded and executed in a standardized manner. The local knowledge of these researchers was vital to ensure the selection of well-informed experts and to minimize the influence of selection bias. Defining experts as those who observe local politics for their profession, four types of experts were interviewed: journalists, academics, NGO activists, and campaign organizers. The selected experts were largely those who had locally acquired a reputation for their knowledge and analysis of local politics. A second selection criterion involved the partisanship of the expert. Using the knowledge of local researchers, we tried to avoid, as much as possible, experts with links to ruling politicians. To neutralize the effect of partisan bias, we tried to make sure that for each expert known to be supportive of the district head, we also included an expert known to be unsupportive. To further ensure the reliability of the results, a local researcher could indicate whenever he or she deemed an interviewed expert to be unreliable. The assessments of these experts were subsequently discarded, involving 24 experts (4.5% of original total of 533 interviewed experts).

This need to work with experienced and locally rooted researchers, affected the selection of districts and provinces. As the number of available researchers per province was limited, I decided to focus on three districts per province, being the provincial capital and two rural districts. Because of this focus on reliability of the implementation, I endeavored to select rural districts for which locally rooted researchers were available while also being, as much as possible, representative in terms of the character of local economy of the province.

The Clientelism Perception Index

This study largely follows an emerging consensus⁸ to define political clientelism (and “clientelistic politics”) as the practice of exchanging a targeted, non-policy-based provision of money and state resources (jobs, public services, government contracts, etc.) for political support (such as votes, campaign funding, and campaign support).⁹ In other words, the exchange between a politician and supporters can be called clientelistic when these supporters receive a personal (i.e., targeted) benefit in a manner perceived to be contingent on his or her support.¹⁰ The defining characteristic of clientelistic politics is thus not the type of benefit that is being offered, or electoral intention behind it, but rather the terms on which this benefit is being offered (Hicken, 2011; Kitschelt, 2000). Clientelistic politics often involves the substitution (or ignoring) of formal criteria and procedures of selecting beneficiaries of government programs for a political criterion: did you, or will you support me? (Hicken, 2011; Stokes, 2007). As this definition implies, political clientelism can involve not only the exchange of various types of personal benefits, ranging from access to state benefits—such as government contracts, public services, or public sector jobs—but also benefits that involve the personal resources of political actors, such as money or protection. Ideally, a measure of the intensity of clientelistic practices encompasses these different forms that clientelistic exchanges may take.

Then what does “more (or less) clientelistic” mean? How do we define degrees of clientelism? The literature is remarkably silent about this conceptual challenge, as most interpretations remain implicit. When “degrees of clientelism” is explicitly conceptualized, the conceptualization is often not comprehensive but rather relates to a proxy for which statistical material is available, such as “a higher ratio of civil servants” (Keefer, 2007) or “more mayoral intervention in selection of welfare recipients” (Weitz-Shapiro, 2014). Herbert Kitschelt’s expert survey offers a more comprehensive and explicit conceptualization, and defines intensity of clientelistic efforts in terms of “how much effort do candidates expend by providing preferential access to” a range of benefits (see Kitschelt et al., 2009). This conceptualization also has drawbacks as it provides no yardstick for interpreting what might constitute a small or large effort. This creates an “anchorage problem”: respondents might have different interpretations of what a big effort might look like.

To address these issues, I adopt a slightly different conceptualization and define degrees of clientelism in terms of the share of state benefits and campaign funds distributed in a manner perceived to be contingent on electoral support. A low degree of clientelism thus implies that hardly any state benefits are perceived to be provided as reward for electoral support and few voters receive monetary incentives, while the highest degree of clientelism signifies that virtually all state benefits are perceived to be provided in exchange for electoral support.

This conceptualization avoids the anchorage problem mentioned above because the survey asked respondents to explicitly assess the proportion of particular state benefits that are provided as a reward for electoral support. For example, the survey asked “In your estimation, of all the major contracts that the district government awards, what percentage goes to companies or businessmen that have supported election campaigns of ruling politicians (bupati, governor, members of dprd) during elections?” Similarly, the survey asked “What percentage of the [higher level] civil servants in the district government have gotten the promotion to this posting as a reward for supporting—openly or secretly—a candidate during elections?” To assess vote buying, the survey used a slightly different yardstick: “What percentage of voters is given money or consumer goods during elections for district head?” When such questions would not be possible, the questions were phrased in terms of the likelihood of a clientelistic distribution: “When implementing welfare programs . . . how likely is it that local government representatives prioritize people who voted for their preferred candidate . . .?” For the formulation of these questions, I leaned on ethnographic fieldwork, trying to ensure that the formulation was both simple and contextually sensitive (see supplementary material for an overview of the survey tool and the implementation of the survey).

To develop the clientelism perceptions index (CPI), I combined measures on seven types of personal benefits (see Table 1) that feature prominently in clientelistic electoral strategies in Indonesia. I selected these seven types on the basis of fieldwork findings and by drawing on Kitschelt’s survey tool (Kitschelt et al., 2009) and on the broader comparative literature on clientelism. The survey asked questions about the distribution of each of these benefits, with regard to both district and provincial government levels and elections.

Table 1.

The Clientelism Perception Index.

Combines assessments about the degree to which the following resources are distributed in a clientelistic manner

1. Governmental contracts (to build a road, supply goods, etc.) 2. Government jobs 3. Public services (preferential access to water, education, sanitation, electricity, etc.) 4. Access to social welfare programs 5. Use of social assistance funds (hibah/bansos) 6. Paperwork (licenses, permits, etc.) 7. Money (vote buying in government head elections)

As Table 2 illustrates, it turned out that, in general, scores on different forms of clientelistic exchanges were highly intercorrelated.¹¹ For example, in areas where government contracts are highly likely to be distributed as rewards to supporters of a successful district head candidate, government jobs are also likely to be distributed in exchange for electoral support. The exception is vote buying during legislative assembly elections (7b). This is an interesting finding to which I return below. Because of this lack of correlation with the other variables, I excluded vote buying during legislative assembly elections from the index variable. The online appendix provides an overview of component scores as well as an assessment of interexpert agreement for these variables. The latter shows that, in general, in high-scoring regions this degree of agreement is relatively high, while for low-scoring regions like Jakarta and Surabaya disagreement among experts is more common.¹² This correlation between CPI scores and interexpert agreement suggests that experts have greater difficulty assessing levels of clientelism in regions where such practices are neither absent nor pervasive. And perhaps recent reform measures adopted by reformist politicians in Java complicated these assessments in such districts. Whatever the reason, this means that the CPI scores are less reliable in many of lower scoring districts. This differential measurement error poses a challenge for regression analysis, which I take up below.

Table 2.

Component Variables Clientelism Perception Index: Correlations of District Averages.

	1	2	3	4	5	6	7a	7b
1. Public services	—
2. Access to welfare	−.214	—
3. Social assistance	−.328**	.466***	—
4. Govt. jobs	−.287	.567***	.811***	—
5. Govt. contracts	−.361**	.601***	.684***	.710***	—
6. Business license	−.479***	.411**	.693***	.687***	.703***	—
7a. Vote buying—district head elections	−.275	.503***	.451***	.416***	.608***	.415***	—
7b. Vote buying—legislative elections	−.289	.14	.182	.156	.410**	.277	.652***	—

The question regarding public services was inversely formulated, hence the negative correlations.

^**p < .05. ***p < .01 (two-tailed).

The standardized scores on all the other component variables were added together to create district-level and provincial-level measures of the extent to which the distribution of state resources is perceived as being contingent on political support in the region concerned. This “Clientelism Perception Index” was recalculated into a scale from 0 to 10. A 0 score implies that no state resources are perceived to be distributed in a manner contingent on electoral support. A score of 10 signifies that all state benefits are seen as being distributed in a clientelistic manner.

An Overview of the Findings¹³

The expert assessments and the clientelism perception index generated from them yield two general conclusions about the nature of clientelistic politics in Indonesia. A first general conclusion is that clientelistic practices are be perceived to be pervasive—but not all-pervasive. Over 70% of all observers, for example, estimated that more than 40% of the big contracts that district government awards are given to campaign supporters (69% thought so about contracts awarded by provincial governments). Most observers (56%) estimated that over 60% of higher level civil servants received their position as a reward for campaign support. Asked about the use of social assistance budgets, 66% of experts estimated that over 60% of these budgets are used to reward campaign supporters.

Yet not all types of clientelistic exchanges are perceived to be equally common. Generally speaking, local observers consider clientelistic exchanges between politicians and their campaign supporters to be more pervasive compared with direct clientelistic exchanges between politicians and voters. Indonesian voters are, it seems, much more likely to be lured through vote buying than through clientelistic distribution of state programs and budgets. Vote buying was considered to be rampant: A large majority of experts (71%) estimated that over 60% of voters received monetary incentives during parliamentary elections (note, however, that this figure is much higher than figures derived from direct survey questions of voters, which generally point to 20% to 30% of Indonesian voters receiving payments at election times (see International Foundation for Election Studies [IFES], 2014). Vote buying was considered slightly less common in district head (49%), gubernatorial (33%), and, particularly, presidential elections (13%). In other words, the more local the electoral contest, the more pervasive the practice of vote buying is perceived to be.

A second general observation is that the perceived intensity of clientelistic practices varies considerably across Indonesia. The scores—ranging from 3.97 (Surabaya) to 7.95 (Kupang) on a scale from 0 to 10—suggest that politics in Java differs considerably from politics in east Indonesia. Figure 1 provides an overview of the average scores of districts, while Figure 2 shows how the studied 16 provincial governments score on the clientelism perception index. These CPI scores reveal a relatively clear pattern, while also throwing up a number of puzzles. The lowest scores are persistently found in Java, particularly its cities like Surabaya, Jakarta (4.2), and Bandung and Tangerang (both 5.0). The Javanese provincial governments, excepting Banten, also score lower than other provinces, with the lowest score for Central Java (4.0). The highest CPI scores are given to districts in Kalimantan and eastern Indonesia, with particularly high scores for district capitals like Kupang (8.0), Makassar (7.4), Palangka Raya (7.4), as well as rural areas in Flores (Manggarai Barat, 7.5), Papua (Jayawijaya, 7.4), and Kalimantan (e.g., Gunung Mas, 7.9, and Kutai Kertanegara, 7.1). The assessment of provincial politics in eastern Indonesia led to similar high scores. On the whole, Sumatran districts and provinces score in between these extremes.

Figure 1.

District-level clientelism perceptions index scores.

Figure 2.

Province-level clientelism perceptions index scores.

A relatively consistent pattern emerges: low scores in West, Central, and East Java, particularly the cities, while experts perceive clientelistic practices to be much more pervasive in eastern Indonesia, including its provincial capitals. Yet within this general pattern, a number of intriguing results are worth highlighting. In Java, the scores largely follow the expected pattern of low scores in cities and slightly higher scores in rural areas. Outside Java, this pattern is more or less reversed as provincial capitals like Kupang, Makassar, Medan, and Palangka Raya score higher than the rural districts nearby. Big and relatively wealthy cities like Medan and Makassar score similar scores as Kalimantan’s poor countryside, while Java’s dense and poverty-prone countryside scores much lower than natural resource–rich rural areas like Bulangan, Gunung Mas, and Kutai Kertanegara. This pattern provides a first indication of the limited applicability of the “cost perspective”: Rural areas in Java, such as Garut, Cianjur, or Batang received low scores despite having high poverty rates and relative low household expenditure. Conversely, relatively affluent provincial capitals such as Makassar, Manado, or Medan as well as rich, natural resource–dependent districts in Kalimantan received high CPI scores.

This pattern largely resonates with descriptive assessments of local politics. Over the last few years, Javanese cities have seen the rise of a different category of local politicians who gained admiration for adopting a different leadership style. Leaders in places such as Surabaya, Bandung, and Solo have become known for their emphasis on good governance and merit-based bureaucratic appointments. These politicians have also been among the first to adopt explicitly programmatic campaign strategies as they start to make policy-based, populist promises about subsidized health care, minimum wage, or free education. In contrast, provinces like Banten, Lampung, Central Kalimantan, and Sulawesi that received high CPI scores are known for harboring entrenched political dynasties whose dominance largely rests on targeted distribution of state resources (Buehler, 2013). Other studies highlight the state-dependent nature of the economy in provincial capitals outside Java, arguing that they grew through a process of “state-sponsored urbanization” (van Klinken, 2014, p. 32). Lacking a broad industrial base, these provincial towns harbor a particular type of middle class consisting of, mainly, civil servants, contractors, and owners of small businesses that are highly dependent on state budgets (see van Klinken & Berenschot, 2014). With so many livelihoods dependent on the state, local politics is described as a heated, faction-ridden affair with an intense involvement of civil servants wanting to safeguard their privileges (Tidey, 2012).¹⁴ Studies that compared the character of politics in different districts (e.g., Djani, 2013; Patunru et al., 2012; Rosser & Wilson, 2012; von Luebke, 2009) yielded assessments that also correspond to the pattern I found.

What Makes Java Less Clientelistic?

How can we explain this pattern? In particular, what is it about West, Central and East Java that make these provinces less prone to clientelistic politics? To answer these questions, I test to what extent the “cost” and “constraint” perspectives summarized above correspond to the empirical findings. I do so by operationalizing these perspectives within the limitations of available data, and subsequently I try to discriminate between them by means of employing a bivariate and multivariate regression framework. To operationalize the cost perspective, I employ poverty rates, years of schooling, household expenditure, and urban population percentage as explanatory variables. To operationalize the constraint perspective, I adopt district government expenditure (as share of total regional GDP),¹⁵ the ratio of jobs in trade, industry, and finance sectors vis-à-vis government jobs (as indicators of state dependency) and the percentage of industry jobs as well as the combined share of the industry, trade, and finance sectors in the district GDP (as indicators of the diversification of local economies).¹⁶ As control variables, I have added population size (a factor known to affect clientelistic strategies, see Brusco et al., 2004), the size of the mining sector, and a Java dummy. Finally, in the regression analyses, I control for unobserved regional effects by systematically applying robust standard errors that are clustered errors per province.¹⁷

The unequal variances in interexpert agreement across districts—with lower degree of interexpert agreement in less clientelistic districts (see above)—violates the assumption of constant variance in the errors that underlies the standard ordinary least squares (OLS) regression approach. To correct for the implied potential bias in OLS regression results, I adopted a weighted-least-square (WLS) regression approach in which the dependent variable is given differential weights across districts. More specifically, I computed the regression with analytic weights proportional to the inverse of the squared standard deviations of the average district-level variance of expert assessments for all seven component variables, such that observations with smaller variance carry larger weight in the regression.¹⁸ In Table 3, I present bivariate regression results to assess the individual explanatory capacity of the abovementioned variables, while Table 4 presents multivariate models.

Table 3.

CPI—Bivariate Models (WLS Regression).

Dependent variable: CPI district	Regression coefficient (SE)	Constant	R 2
LN_household expenditure	−0.582	13.927	.06
	(1.56)	(2.81)**
Poverty rate	0.042	5.566	.12
	(3.03)***	(18.71)***
Years of schooling	−0.126	7.194	.05
	(1.52)	(10.45)***
Urban population share	−0.009	6.585	.11
	(2.03)*	(29.25)***
Industry Jobs, share of total	−0.063	6.742	.27
	(3.62)***	(27.77)***
Ratio jobs industry/civil servants	−0.096	6.503	.33
	(2.88)**	(32.86)***
Ratio jobs industry, trade, finance/civil servants	−0.035	6.696	.40
	(4.48)***	(29.60)***
Relative size industry, trade, finance sectors	−0.022	6.961	.22
	(2.77)**	(20.11)***
Share government expenditure in district GDP	0.029	5.693	.19
	(2.49)**	(21.69)***

See the online appendix for sources of the data used. N = 38. CPI = clientelism perception index; WLS = weighted least squares.

*p < .1. **p < .05. ***p < .01.

Table 4.

CPI—Multivariate Models With Controls (WLS Regression).

Dependent variable: CPI	4a “cost”	4b “constraint”	4c mixed	4d with controls	4e with provincial effects
LN_household expenditure	0.278		0.043	0.057	−0.957
	(0.30)		(0.07)	(0.08)	(0.86)
Poverty rate	0.034		0.006	0.013	0.046
	(1.48)		(0.21)	(0.56)	(1.90)*
Years of schooling	0.144		−0.054	0.087	0.730
	(0.66)		(0.23)	(0.38)	(2.32)**
Share urban population	−0.013		0.005	0.004	−0.011
	(1.27)		(0.54)	(0.44)	(0.91)
Share industry jobs		−0.020	−0.016	0.018	0.066
		(0.81)	(0.41)	(0.54)	(1.36)
Ratio jobs industry, trade, finance/civil servants		−0.035	−0.037	−0.064	−0.070
		(4.21)***	(2.91)**	(4.02)***	(3.47)***
Share government expenditure in district GDP		0.012	0.010	0.025	0.021
		(0.94)	(0.67)	(1.35)	(1.69)
Relative size industry, trade, finance sectors		0.013	0.009	0.008	0.010
		(1.35)	(0.66)	(0.51)	(0.62)
LN_population				0.518	0.540
				(1.96)*	(2.33)**
Relative size mining sector				0.025	0.044
				(0.36)	(0.42)
Java dummy				−0.807	−1.859
				(1.23)	(1.42)
Provincial fixed effects	no	no	no	no	yes
_constant	1.378	6.238	5.962	−2.174	6.875
	(0.12)	(11.67)***	(0.72)	(0.23)	(0.44)
R 2	.18	.44	.45	.61	.78
N	38	38	38	35	35

See the online appendix for sources of the data used. CPI = clientelism perceptions index; WLS = weighted least squares.

^*p < .1. **p < .05. ***p < .01.

Given the limited number of observations (districts) included in this analysis, these regression results need to be interpreted with considerable caution. Yet a number of conclusions stand out relatively clearly. Of the variables associated with a cost perspective, only poverty rate and, to a limited extent, urban population share turn out to be significant in the bivariate models. These variables all turn out to be insignificant in the multivariate models, with the partial exception of the model that controls for provincial fixed effects (where, contrary to what one would expect, years of schooling suddenly turns out to be positively related to CPI). The variables associated with the “constraint perspective”—that is, those related to state dependency and economic diversification—generally turn out to have more explanatory power. In the bivariate regression framework, all these variables turn up as significant. The multivariate model associated with the constraint perspective (Model 3b) turns out to have a much stronger predictive capacity compared with the model associated with the cost perspective (3a). In the combined model (3c), particularly the ratio of jobs in industry, trade, and finance versus government jobs turns out to have the strongest impact on the CPI level across districts, a result that appears to be robust when I add control variables (Model 3d) and control for provincial fixed effects (3e). In short, while some variables associated with the “cost perspective” appear as significant in some model specifications, in general these variables are poor predictors of perceived levels of clientelism across Indonesia. In contrast, throughout these different model specifications, the variables associated with the “constraint perspective” constantly have a stronger explanatory value.

While the small number of provinces does not allow for a meaningful regression analysis of provincial CPI scores, Figure 3 suggests that a similar pattern might be found if we look at provincial CPI scores. This figure plots the average of provincial and district CPI scores in relation to the relative size of industry, trade, and finance sectors.

Figure 3.

Average provincial CPI scores and share of industry, trade, and finance sectors.

NTT = Nusa Tenggara Timur; CPI = clientelism perception index.

However, there is one clear exception to this general pattern. As Table 2 illustrated, perceptions of vote buying in, particularly, legislative elections do not strongly correlate with other aspects of clientelistic politics. Such vote buying is perceived to be pervasive throughout Indonesia. Levels of vote buying in elections in rural and indeed in some urban districts in Java are much higher than their scores on other dimensions of the CPI. Javanese districts like Surabaya and Batang are actually among the high scorers in terms of vote buying during legislative assembly elections.

This finding suggests that the relatively high levels of poverty among voters in these regions of Java might make them especially likely to be targeted by vote buying. In other words, the widely agreed consensus in the literature—that higher poverty rates and lower levels of economic development explain electoral clientelism—might hold true for vote buying while being less able to explain the other dimensions of clientelism covered by the clientelism perception index. That vote buying might operate according to a logic is at least in part delinked from other types of clientelistic exchanges that bind politicians to brokers, civil servants, and other elites. This finding suggests that vote buying cannot be taken as a proxy for other types of clientelistic exchange. This underscores the need for employing more comprehensive measures of clientelism such as the CPI index developed in this article.

Conclusion

A major obstacle impeding the comparative study of clientelistic politics is the difficulty of assessing the extent to which clientelistic practices are pervasive. The lack of such comparative measures has particularly hampered the identification of the kinds of conditions that foster clientelistic politics. In this article, I have addressed this challenge by employing a combination of ethnographic fieldwork and an expert survey to compare the nature of local politics in 38 districts in Indonesia. The assessments of these local political observers yielded a relatively consistent, yet intriguing, pattern, with low clientelism perception index scores in Javanese provinces (especially its cities) and high scores in Kalimantan and eastern Indonesia. These findings were relatively consistent in the sense that CPI scores within regions were generally similar, while the contrasts between regions were relatively stark. Furthermore, the findings generally corresponded with my own fieldwork, as well as other qualitative studies, providing a measure of confidence in the results of the expert survey and the general pattern that it identified.

Surprisingly, the dominant “cost perspective” found in the literature on clientelism—focusing on poverty, urbanization, and the size of the middle class—turns out to be of limited use to explain variation across Indonesia. Both the findings from the expert survey as well as my ethnographic fieldwork on election campaigns suggest that it is not so much the degree but rather the character of economic growth that matters. CPI scores were high in areas with extractive economies, as well as in areas with state-dependent economies. The lower scores in Java suggest that economic diversification and a wider distribution of economic power can serve to deter clientelistic practices. On the basis of these findings, I developed an alternative interpretation of the relationship between economic development and clientelistic politics. I have argued that a dispersion of economic power can curtail clientelistic politics because such dispersion can generate a more open public sphere and a more autonomous civil society capable of scrutinizing and disciplining the behavior of politico-business elites. I operationalized this “constraint perspective” by employing a number of variables related to state dependency and economic diversification. Regressing these variables against CPI scores, these variables generally turned out to have much stronger explanatory capacity.

These findings also suggest that the emphasis in literature on conditions favoring clientelism suffers from an overly narrow focus on vote buying. The dominance of the “cost perspective” in the literature on clientelism is partly due to a reliance on general population surveys on vote buying. Lacking alternative research methods, these analyses are not based on assessments of the pervasiveness of other dimensions of clientelistic politics—such as providing government jobs and contracts or access to public services. This article’s findings suggest that the factors and mechanisms driving vote buying might be quite different from those driving the clientelistic distribution of state resources to brokers, civil servants, and business elites. To explain the latter dimensions of clientelistic politics, the concentration of economic control and the associated distribution of power are more relevant. The findings thus highlight the importance of paying attention to different dimensions of clientelistic politics beyond only vote buying. The expert survey instrument employed in this study offers a new avenue to engage in such a more comprehensive comparative study.

Author Biography

Ward Berenschot is a researcher at the Royal Netherlands Institute for Southeast Asian and Caribbean Studies (KITLV), working on democracy and identity politics in India and Indonesia. He is the author of Riot Politics: Hindu-Muslim Violence and the Indian State (Columbia University Press, 2011).